(19)
(11)EP 3 699 817 A2

(12)EUROPEAN PATENT APPLICATION

(43)Date of publication:
26.08.2020 Bulletin 2020/35

(21)Application number: 20169859.4

(22)Date of filing:  16.04.2015
(51)International Patent Classification (IPC): 
G06K 9/00(2006.01)
(84)Designated Contracting States:
AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

(30)Priority: 09.05.2014 KR 20140055687
24.06.2014 KR 20140077333

(62)Application number of the earlier application in accordance with Art. 76 EPC:
15163789.9 / 2942736

(71)Applicant: Samsung Electronics Co., Ltd.
Gyeonggi-do 16677 (KR)

(72)Inventors:
  • KIM, Won Jun
    Suwon-si Gyeonggi-do 443-803 (KR)
  • SUH, Sung Joo
    Suwon-si Gyeonggi-do 443-803 (KR)
  • HAN, Jae Joon
    Suwon-si Gyeonggi-do 443-803 (KR)
  • HWANG, Won Jun
    Suwon-si Gyeonggi-do 443-803 (KR)

(74)Representative: Arnold & Siedsma 
Bezuidenhoutseweg 57
2594 AC The Hague
2594 AC The Hague (NL)

 
Remarks:
This application was filed on 16-04-2020 as a divisional application to the application mentioned under INID code 62.
 


(54)LIVENESS TESTING METHODS AND APPARATUSES AND IMAGE PROCESSING METHODS AND APPARATUSES


(57) An image processing method comprising: receiving a first image including an illumination component and a non-illumination component; filtering a plurality of pixels included in the first image to generate a second image related to the illumination component; and generating a third image related to the non-illumination component based on the first image and the second image.




Description


[0001] One or more example embodiments relate to liveness testing methods, liveness testing apparatuses, image processing methods, image processing apparatuses, and/or electronic devices including the same.

[0002] Biometric technologies may identify a human based on unique biometric characteristics of each individual user. Among conventional biometric technologies, a face recognition system may naturally recognize a user based on the user's face without requiring user contact with a sensor, such as a fingerprint scanner or the like. However, conventional face recognition systems may be vulnerable to impersonations using a picture of a face of a registered target.

[0003] At least one example embodiment provides a liveness testing method including: testing a liveness of an object included in a received input image based on whether an image of the object has a characteristic indicative of a flat surface or a characteristic indicative of a three-dimensional (3D) structure.

[0004] Such embodiments have the advantage of obtaining a reliable result with respect to the determining of liveliness whilst requiring reduced computational complexity compared with prior art methods.

[0005] The image of the object included in the received input image may correspond to a face. In that way it is made possible to distinguish a live face from a picture.

[0006] According to at least some example embodiments, the method may further include: determining whether the image of the object has the characteristic indicative of the flat surface or the characteristic indicative of the 3D structure based on a distribution of light energy included in a plurality of pixels corresponding to the image of the object. Such embodiments are based on the inventive insight that using a light energy distribution, the liveliness can be easily verified without the need for complex computations.

[0007] According to at least some example embodiments, the method may further include: determining whether the image of the object has the characteristic indicative of the flat surface or the characteristic indicative of the 3D structure based on a degree of uniformity in a distribution of light energy included in a plurality of pixels corresponding to the image of the object.

[0008] According to at least some example embodiments, the method may further include: determining whether the image of the object has the characteristic indicative of the flat surface or the characteristic indicative of the 3D structure based on statistical information related to diffusion speeds of a plurality of pixels corresponding to the image of the object. Such embodiments will allow determining liveliness in an improved manner in various lighting conditions.

[0009] According to at least some example embodiments, the method may further include: iteratively calculating values of the plurality of pixels based on a diffusion equation; and calculating the diffusion speeds for each of the plurality of pixels based on a difference between a pixel value before each iterative calculation and a pixel value after each iterative calculation. The statistical information related to the diffusion speeds may include at least one of: a number of pixels, among the plurality of pixels, having diffusion speeds greater than or equal to a threshold value; a distribution of the pixels having the diffusion speeds greater than or equal to the threshold value; an amount of noise components included in a first scale region extracted based on magnitudes of the diffusion speeds; an average of the diffusion speeds; a standard deviation of the diffusion speeds; and a filter response based on the diffusion speeds.

[0010] According to at least some example embodiments, the method may further include: filtering the received input image to generate a filtered image; and determining whether the image of the object has the characteristic indicative of the flat surface or the characteristic indicative of the 3D structure based on statistical information related to a change in values of a plurality of pixels corresponding to the image of the object in the received input image and a plurality of pixels corresponding to the image of the object in the filtered image.

[0011] The filtering may include: diffusing the received input image to generate a diffusion image; calculating diffusion speeds for each of the plurality of pixels corresponding to the image of the object based on a difference between a value of each pixel in the received input image and a corresponding value of the pixel in the diffusion image; and wherein the determining determines whether the image of the object has the characteristic indicative of the flat surface or the characteristic indicative of the 3D structure based on the calculated diffusion speeds.

[0012] According to at least some example embodiments, the method may further include at least one of: outputting a signal corresponding to a failed test when the object is determined to have the characteristic indicative of the flat surface; and outputting a signal corresponding to a successful test when the object is determined to have the characteristic indicative of the 3D structure.

[0013] The input image may be a single image including face, for example.

[0014] At least one other example embodiment provides a liveness testing method including: filtering a received image including an image of an object to generate a filtered image; determining a magnitude of change in values of pixels corresponding to the image of the object in the filtered image and the received image; and testing a liveness of the object based on the magnitude of change. Such embodiments have the advantage of obtaining a reliable result with respect to the determining of liveliness using simple filtering techniques.

[0015] The filtering may include diffusing the pixels corresponding to the image of the object in the received image to generate a diffusion image, and wherein the method may further include: calculating diffusion speeds for pixels corresponding to the image of the object based on the values of the pixels corresponding to the image of the object in the received image and the diffusion image; and wherein the testing tests the liveness of the object based on the calculated diffusions speeds.

[0016] In example embodiments the filtering may include bilateral filtering. More generally, any suitable filtering that preserves an edge region and blurs a non-edge region, typically in a manner similar or substantially similar to diffusion and bilateral filtering, may be used in connection with example embodiments discussed herein.

[0017] According to at least some example embodiments, the diffusing may further include: iteratively updating values of the pixels based on a diffusion equation. The iteratively updating may iteratively update the values of the pixels by applying an additive operator splitting (AOS) scheme to the diffusion equation. Such an embodiment may result in the generation of an illumination-invariant image of the object which may be applicable for object recognition.

[0018] According to at least some example embodiments, the testing may include: estimating a surface property related to the object based on the diffusion speeds; and testing the liveness of the object based on the estimated surface property.

[0019] The surface property may include at least one of: a light-reflective property of a surface of the object; a number of dimensions of the surface of the object; and a material of the surface of the object.

[0020] According to at least some example embodiments, the estimating may include: analyzing a distribution of light energy included in the image of the object based on the diffusion speeds to estimate the surface property.

[0021] According to at least some example embodiments, the method may further include at least one of: outputting a signal corresponding to a failed test when the estimated surface property corresponds to a surface property of a medium displaying a face; and outputting a signal corresponding to a successful test when the estimated surface property corresponds to a surface property of an actual face.

[0022] According to at least some example embodiments, the testing may further include: calculating statistical information related to the diffusion speeds; and testing the liveness of the object based on the calculated statistical information.

[0023] According to at least some example embodiments, the calculating statistical information may include at least one of: calculating a number of pixels having diffusion speeds greater than or equal to a threshold value, among the diffusion speeds; calculating a distribution of the pixels having the diffusion speeds greater than or equal to the threshold value, among the diffusion speeds; calculating at least one of an average and a standard deviation of the diffusion speeds; and calculating a filter response based on the diffusion speeds.

[0024] According to at least some example embodiments, the calculating statistical information may further include: extracting a first scale region from the received image based on magnitudes of the diffusion speeds; extracting a characteristic of the first scale region; and wherein the testing tests the liveness of the object based on the extracted characteristic.

[0025] The characteristic of the first scale region may include an amount of noise components included in the first scale region, and the noise components may be calculated based on a difference between the first scale region and a result of applying median filtering to the first scale region.

[0026] According to at least some example embodiments, the method may further include: outputting a signal corresponding to a failed test when the statistical information corresponds to statistical information related to a medium displaying a face; and outputting a signal corresponding to a successful test when the statistical information corresponds to statistical information related to an actual face.

[0027] According to at least some example embodiments, the calculating the diffusion speeds may include: calculating a diffusion speed for each of the pixels based on an original value of the pixel before diffusion and a diffusion value of the pixel after diffusion. The calculated diffusion speed of the pixel increases as the difference between the original value and the diffusion value increases, and the calculated diffusion speed of the pixel decreases as the difference between the original value and the diffusion value decreases.

[0028] At least one other example embodiment provides an image processing method including: receiving a first image including an illumination component and a non-illumination component; filtering a plurality of pixels included in the first image to generate a second image related to the illumination component; and generating a third image related to the non-illumination component based on the first image and the second image. Such a third image will allow performing object recognition in an improved manner.

[0029] The second image may be a diffusion image, and the filtering may include: diffusing the plurality of pixels included in the first image to generate the diffusion image.

[0030] According to at least some example embodiments, the method may further include at least one of: recognizing a face based on the third image; and verifying a user based on the third image.

[0031] The diffusing may include: iteratively updating values of the plurality of pixels by applying an additive operator splitting (AOS) scheme to a diffusion equation.

[0032] The third image may be generated based on at least one of a ratio of the first image to the second image and a difference between the first image and the second image in a log domain.

[0033] The non-illumination component may be included in a first scale region impervious to a change in illumination, and the illumination component may be included in a second scale region sensitive to a change in illumination.

[0034] A diffusion speed of a pixel corresponding to the non-illumination component may be higher than a diffusion speed of a pixel corresponding to the illumination component.

[0035] At least one other example embodiment provides a non-transitory computer-readable medium comprising a program that, when executed on a computer device, causes the computer device to perform any one of the example embodiments of a method disclosed above.

[0036] At least one other example embodiment provides a liveness testing apparatus including: a testing circuit configured to test a liveness of an object included in a received input image based on whether an image of the object has a characteristic indicative of a flat surface or a characteristic indicative of a three-dimensional (3D) structure.

[0037] The testing circuit may be further configured to: filter the received input image to generate a filtered image; and determine whether the image of the object has the characteristic indicative of the flat surface or the characteristic indicative of the 3D structure based on statistical information related to a change in values of a plurality of pixels corresponding to the image of the object in the received input image and a plurality of pixels corresponding to the image of the object in the filtered image.

[0038] The testing circuit may be further configured to: diffuse the received input image to generate a diffusion image; calculate diffusion speeds for each of the plurality of pixels corresponding to the image of the object based on a difference between a value of each pixel in the received input image and a corresponding value of the pixel in the diffusion image; and determine whether the image of the object has the characteristic indicative of the flat surface or the characteristic indicative of the 3D structure based on the calculated diffusion speeds.

[0039] At least one other example embodiment provides an image processing apparatus including: a filter circuit configured to receive a first image including an illumination component and a non-illumination component; a filter circuit configured to filter a plurality of pixels included in the first image to generate a second image related to the illumination component; and a generator circuit configured to generate a third image related to the non-illumination component based on the first image and the second image.

[0040] According to another aspect of the invention there is provided a method according to any one of the following clauses:

1. A liveness testing method comprising:

filtering a received image including an image of an object to generate a filtered image;

determining a magnitude of change in values of pixels corresponding to the image of the object in the filtered image and the received image; and

testing a liveness of the object based on the magnitude of change.

2. The liveness testing method of clause 1, wherein the filtering includes
diffusing the pixels corresponding to the image of the object in the received image to generate a diffusion image, and wherein the method further includes,
calculating diffusion speeds for pixels corresponding to the image of the object based on the values of the pixels corresponding to the image of the object in the received image and the diffusion image, and wherein
the testing tests the liveness of the object based on the calculated diffusions speeds.

3. The method of clause 1 or 2, wherein the object corresponds to a face.

4. The method of clause 2, wherein the diffusing comprises:

iteratively updating values of the pixels based on a diffusion equation.

5. The method of clause 4, wherein the iteratively updating iteratively updates the values of the pixels by applying an additive operator splitting (AOS) scheme to the diffusion equation.

6. The method of any one of the previous clauses, wherein the testing comprises:

estimating a surface property related to the object based on the diffusion speeds; and

testing the liveness of the object based on the estimated surface property.

7. The method of clause 6, wherein the surface property comprises at least one of:

a light-reflective property of a surface of the object;

a number of dimensions of the surface of the object; and

a material of the surface of the object.

8. The method of clause 7, wherein the estimating comprises:
analyzing a distribution of light energy included in the image of the object based on the diffusion speeds to estimate the surface property.

9. The method of any one of the previous clauses, further comprising at least one of:

outputting a signal corresponding to a failed test when the estimated surface property corresponds to a surface property of a medium displaying a face; and

outputting a signal corresponding to a successful test when the estimated surface property corresponds to a surface property of an actual face.

10. The method of clause 2, wherein the testing comprises:

calculating statistical information related to the diffusion speeds; and

testing the liveness of the object based on the calculated statistical information.

11. The method of clause 10, wherein the calculating statistical information comprises at least one of:

calculating a number of pixels having diffusion speeds greater than or equal to a threshold value, among the diffusion speeds;

calculating a distribution of the pixels having the diffusion speeds greater than or equal to the threshold value, among the diffusion speeds;

calculating at least one of an average and a standard deviation of the diffusion speeds; and

calculating a filter response based on the diffusion speeds.

12. The method of clause 11, wherein the calculating statistical information comprises:

extracting a first scale region from the received image based on magnitudes of the diffusion speeds; and

extracting a characteristic of the first scale region; and wherein

the testing tests the liveness of the object based on the extracted characteristic.

13. The method of clause 12, wherein
the characteristic of the first scale region includes an amount of noise components included in the first scale region; and
the noise components are calculated based on a difference between the first scale region and a result of applying median filtering to the first scale region.

14. The method of any one of the clauses 10-13, further comprising:

outputting a signal corresponding to a failed test when the statistical information corresponds to statistical information related to a medium displaying a face; and

outputting a signal corresponding to a successful test when the statistical information corresponds to statistical information related to an actual face.

15. The method of any one of the clauses 10-14, wherein the calculating the diffusion speeds comprises:
calculating a diffusion speed for each of the pixels based on an original value of the pixel before diffusion and a diffusion value of the pixel after diffusion.

16. The method of clause 15, wherein the calculated diffusion speed of the pixel increases as the difference between the original value and the diffusion value increases, and the calculated diffusion speed of the pixel decreases as the difference between the original value and the diffusion value decreases.

17. The method of any one of the previous clauses, wherein the input image corresponds to a single image of a face of a user.

9. A non-transitory computer-readable medium comprising a program that, when executed on a computer device, causes the computer device to perform the method of any one of the previous clauses.



[0041] According to another aspect of the invention there is provided a method according to any one of the following clauses:
  1. 1. An image processing method comprising:

    receiving a first image including an illumination component and a non-illumination component;

    filtering a plurality of pixels included in the first image to generate a second image related to the illumination component; and

    generating a third image related to the non-illumination component based on the first image and the second image.

  2. 2. The method of clause 1, wherein the second image is a diffusion image, and the filtering comprises:
    diffusing the plurality of pixels included in the first image to generate the diffusion image.
  3. 3. The method of clause 2, wherein the diffusing comprises:
    iteratively updating values of the plurality of pixels by applying an additive operator splitting (AOS) scheme to a diffusion equation.
  4. 4. The method of claim 2 or 3, wherein a diffusion speed of a pixel corresponding to the non-illumination component is higher than a diffusion speed of a pixel corresponding to the illumination component.
  5. 5.The method of any one of the previous clauses, further comprising at least one of:

    recognizing a face based on the third image; and

    verifying a user based on the third image.

  6. 6. The method of any one of the previous clauses, wherein the third image is generated based on at least one of a ratio of the first image to the second image and a difference between the first image and the second image in a log domain.
  7. 7. The method of any one of the previous clauses, wherein the non-illumination component is included in a first scale region impervious to a change in illumination, and the illumination component is included in a second scale region sensitive to a change in illumination.
  8. 8. The method of any one of the previous clauses, wherein the first image corresponds to a single image of a face.
  9. 9. A non-transitory computer-readable medium comprising a program that, when executed on a computer device, causes the computer device to perform the method of any one of the previous clauses.


[0042] According to another aspect of the invention there is provided a method according to any one of the following clauses:
  1. 1. An image processing apparatus comprising:

    a filter circuit configured to receive a first image including an illumination component and a non-illumination component;

    a filter circuit configured to filter a plurality of pixels included in the first image to generate a second image related to the illumination component; and

    a generator circuit configured to generate a third image related to the non-illumination component based on the first image and the second image.

  2. 2. The apparatus of clause 1, wherein the first image corresponds to a single image including a face.

BRIEF DESCRIPTION OF THE DRAWINGS



[0043] Example embodiments will become more apparent and more readily appreciated from the following description of the example embodiments shown in the drawings in which:

FIGS. 1A and 1B illustrate a liveness test according to example embodiments;

FIG. 2 illustrates a principle of a liveness test according to an example embodiment;

FIG. 3 illustrates a liveness testing apparatus according to an example embodiment;

FIG. 4 illustrates a diffusion process according to an example embodiment;

FIG. 5 illustrates an example small-scale region (SR) map according to example embodiments;

FIG. 6 illustrates a liveness testing apparatus according to an example embodiment;

FIG. 7 illustrates an example input image and example images processed according to an example embodiment;

FIG. 8 illustrates example changes in an input image as a result of changes in illumination, according to example embodiments;

FIG. 9 illustrates an image processing apparatus according to an example embodiment;

FIG. 10 illustrates a liveness testing method according to an example embodiment;

FIG. 11 illustrates an image processing method according to an example embodiment;

FIG. 12 illustrates an image processing and authentication/verification method according to another example embodiment; and

FIG. 13 is a block diagram illustrating an electronic system according to an example embodiment.


DETAILED DESCRIPTION



[0044] Various example embodiments will now be described more fully with reference to the accompanying drawings in which some example embodiments are shown.

[0045] Detailed illustrative embodiments are disclosed herein. However, specific structural and functional details disclosed herein are merely representative for purposes of describing example embodiments. Example embodiments may, however, be embodied in many alternate forms and should not be construed as limited to only the embodiments set forth herein.

[0046] Accordingly, while example embodiments are capable of various modifications and alternative forms, the embodiments are shown by way of example in the drawings and will be described herein in detail. It should be understood, however, that there is no intent to limit example embodiments to the particular forms disclosed. On the contrary, example embodiments are to cover all modifications, equivalents, and alternatives falling within the scope of this disclosure. Like numbers refer to like elements throughout the description of the figures.

[0047] Although the terms first, second, etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first element could be termed a second element, and similarly, a second element could be termed a first element, without departing from the scope of this disclosure. As used herein, the term "and/or," includes any and all combinations of one or more of the associated listed items.

[0048] When an element is referred to as being "connected," or "coupled," to another element, it can be directly connected or coupled to the other element or intervening elements may be present. By contrast, when an element is referred to as being "directly connected," or "directly coupled," to another element, there are no intervening elements present. Other words used to describe the relationship between elements should be interpreted in a like fashion (e.g., "between," versus "directly between," "adjacent," versus "directly adjacent," etc.).

[0049] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting. As used herein, the singular forms "a," "an," and "the," are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises," "comprising," "includes," and/or "including," when used herein, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.

[0050] It should also be noted that in some alternative implementations, the functions/acts noted may occur out of the order noted in the figures. For example, two figures shown in succession may in fact be executed substantially concurrently or may sometimes be executed in the reverse order, depending upon the functionality/acts involved.

[0051] Specific details are provided in the following description to provide a thorough understanding of example embodiments. However, it will be understood by one of ordinary skill in the art that example embodiments may be practiced without these specific details. For example, systems may be shown in block diagrams so as not to obscure the example embodiments in unnecessary detail. In other instances, well-known processes, structures and techniques may be shown without unnecessary detail in order to avoid obscuring example embodiments.

[0052] In the following description, example embodiments will be described with reference to acts and symbolic representations of operations (e.g., in the form of flow charts, flow diagrams, data flow diagrams, structure diagrams, block diagrams, etc.) that may be implemented as program modules or functional processes include routines, programs, objects, components, data structures, etc., that perform particular tasks or implement particular abstract data types and may be implemented using existing hardware at, for example, existing electronic devices, such as smartphones, personal digital assistants, laptop or tablet computers, etc. Such existing hardware may include one or more Central Processing Units (CPUs), graphics processing units (GPUs), image processors, system-on-chip (SOC) devices, digital signal processors (DSPs), application-specific-integrated-circuits, field programmable gate arrays (FPGAs) computers or the like.

[0053] Although a flow chart may describe the operations as a sequential process, many of the operations may be performed in parallel, concurrently or simultaneously. In addition, the order of the operations may be re-arranged. A process may be terminated when its operations are completed, but may also have additional steps not included in the figure. A process may correspond to a method, function, procedure, subroutine, subprogram, etc. When a process corresponds to a function, its termination may correspond to a return of the function to the calling function or the main function.

[0054] Reference will now be made in detail to the example embodiments illustrated in the accompanying drawings, wherein like reference numerals refer to like elements throughout. Example embodiments are described below to explain the present disclosure by referring to the figures. One or more example embodiments described below may be applicable to various fields, for example, smartphones, laptop or tablet computers, smart televisions (TVs), smart home systems, smart cars, surveillance systems, etc. For example, one or more example embodiments may be used to test a liveness of an input image and/or authenticate a user to log into a smartphone or other device. In addition, one or more example embodiments may be used to test a liveness of an input image and/or authenticate a user for admission control and/or monitoring in a public area and/or a secured area.

Liveness test according to example embodiments



[0055] FIGS. 1A and 1B illustrate a liveness test according to an example embodiment.

[0056] According to at least some example embodiments, a liveness test refers to a method of testing (or determining) whether an object included in an input image corresponds to a real three dimensional object. In one example, the liveness test may verify whether a face included in an input image corresponds to (or is obtained from) a real three-dimensional (3D) object (e.g., an actual face) or a fake two-dimensional (2D) representation of the object (e.g., a picture of a face). Through the liveness test, an attempt to verify a face of another using a forged and/or falsified picture may be effectively rejected.

[0057] Referring to FIGS. 1A and 1B, according to at least one example embodiment, a liveness testing apparatus 110 receives an input image including a face of a user 120, and tests a liveness of the face included in the received input image. In one example, the liveness testing apparatus 110 may be (or be included in) a mobile device, for example, a mobile phone, a smartphone, a personal digital assistant (PDA), a tablet computer, a laptop computer, etc. In another example, the liveness testing apparatus 110 may be (or be included in) a computing device, for example, a personal computer (PC), an electronic product, such as a TV, a security device for a gate control, etc. The liveness testing apparatus 110 may receive the input image from an image sensor 115 that photographs the face of the user 120. The image sensor 115 may also be part of the liveness testing apparatus 110.

[0058] In one example, as shown in FIG. 1A, the input image is generated by photographing the actual face of the user 120. In this example, the liveness testing apparatus 110 determines that the face included in the input image corresponds to a real (or living) three-dimensional object, and outputs a signal indicating that the face included in the input image corresponds to a real three-dimensional object. That is, for example, the liveness testing apparatus 110 tests whether the face included in the input image corresponds to the real (or living) three-dimensional object, and outputs a signal indicating a successful test since the face in the input image does indeed correspond to the real (or living) three-dimensional object.

[0059] In another example, as shown in FIG. 1B, the input image is generated by photographing a face displayed on a display medium 125, rather than the actual face of the user 120. According to at least this example, the display medium 125 refers to a medium that displays an object (e.g., a face) in two dimensions. The display medium 125 may include, for example, a piece of paper on which a face of a user is printed (e.g., a photo), an electronic device displaying a face of a user, etc. In one example scenario, the user 120 may attempt to log into an electronic device (e.g., a smartphone or the like) with another user's account by directing a face displayed on the display medium 125 toward the image sensor 115. In FIG. 1B, the face displayed on the display medium 125 is marked with a broken line to indicate that the face displayed on the display medium 125 is directed toward the image sensor 115, rather than the user 120. In this example, the liveness testing apparatus 110 determines that the face included in the input image corresponds to a fake two-dimensional representation of the object, and outputs a signal indicating that the face included in the input image corresponds to a fake two-dimensional representation of the object. That is, for example, the liveness testing apparatus 110 tests whether the face included in the input image corresponds to the real (or living) three-dimensional object, and outputs a signal indicating a failed test since the face in the input image does not correspond to the real (or living) three-dimensional object, but rather the fake two-dimensional representation of the object. In some cases, the term fake object may be used to refer to the fake two-dimensional representation of the object.

[0060] The liveness testing apparatus 110 may detect a face region from the input image. In this example, liveness testing method and apparatus may be applicable to the face region detected from the input image.

[0061] FIG. 2 illustrates a principle of a liveness test according to an example embodiment.

[0062] A liveness testing apparatus according to at least this example embodiment tests a liveness of an object included in an input image based on whether the object has one or more characteristics of a flat (two-dimensional) surface or a three-dimensional (3D) structure.

[0063] Referring to FIG. 2, the liveness testing apparatus distinguishes between a face 211 displayed on a medium 210 and an actual face 220 of a user. The face 211 displayed on the medium 210 corresponds to a two-dimensional (2D) flat surface. When an input image is generated by photographing the face 211 displayed on the medium 210, an object included in the input image has one or more characteristics of a flat surface. Since a surface of the medium 210 corresponds to a 2D flat surface, light 215 incident on the face 211 displayed on the medium 210 is more uniformly reflected by the surface of the medium 210. Thus, light energy is more uniformly distributed on the object included in the input image. Even if the medium 210 is curved, a surface of a curved display may still correspond to, and have characteristics, of a 2D flat surface.

[0064] Conversely, the actual face 220 of the user is a 3D structure. When an input image is generated by photographing the actual face 220 of the user, an object included in the input image has characteristics of a 3D structure. Since the actual face 220 of the user corresponds to a 3D structure having various 3D curves and shapes, light 225 incident to the actual face 220 of the user is less (or non-) uniformly reflected by a surface of the actual face 220 of the user. Thus, light energy is less (or non-) uniformly distributed on the object included in the input image.

[0065] According to at least one example embodiment, a liveness testing apparatus tests a liveness of an object included in an input image based on a distribution of light energy in the object. In one example, the liveness testing apparatus analyzes the distribution of light energy included in the object of the input image to determine whether the object included in the input image has characteristics of a flat 2D surface or of a 3D structure.

[0066] Still referring to FIG. 2, in one example, an input image may be generated by photographing the face 211 displayed on the medium 210. In this case, the liveness testing apparatus analyzes a distribution of light energy included in the face of the input image, and determines that the face of the input image has one or more characteristics of a flat surface. The liveness testing apparatus then determines that the face of the input image corresponds to a fake object, and outputs a signal indicating the same (e.g., indicating a failed test).

[0067] In another example, an input image may be generated by photographing the actual face 220 of the user. In this case, the liveness testing apparatus analyzes a distribution of light energy included in a face of the input image, and determines that the face of the input image has one or more characteristics of a 3D structure. The liveness testing apparatus then determines that the face of the input image corresponds to a real 3D object, and outputs a signal indicating the same (e.g., indicating a successful test).

[0068] According to at least some example embodiments, the liveness testing apparatus may determine a liveness of an object included in an input image based on a degree of uniformity in the distribution of light energy included in the object in the input image. With regard again to FIG. 2, in one example, since the light 215 incident to the face 211 displayed on the medium 210 is reflected substantially uniformly, light energy included in the face of the input image is distributed substantially uniformly. When a degree of uniformity in the distribution of the light energy included in the face of the input image is greater than or equal to a given (or alternatively, desired or predetermined) threshold degree of uniformity, the liveness testing apparatus determines that the face of the input image has one or more characteristics of a flat 2D surface. In this case, the liveness testing apparatus determines that the face of the input image corresponds to a fake object, and outputs a signal indicating the same (e.g., indicating a failed test).

[0069] In another example with regard to FIG. 2, since the light 225 incident on the actual face 220 of the user is less (or non-) uniformly reflected, light energy included in the face of the input image has a light distribution that is less (or non-) uniform. When a degree of uniformity in the distribution of the light energy included in the face of the input image is less than the given threshold degree of uniformity, the liveness testing apparatus determines that the face of the input image has one or more characteristics of a 3D structure. In this case, the liveness testing apparatus determines that the face of the input image corresponds to a real 3D object, and outputs a signal indicating the same (e.g., indicating a successful test).

[0070] In one example, the threshold degree of uniformity may be a value corresponding to a situation in which about 50% or more of a number of pixels included in an image portion correspond to a face region (e.g., more generally, a region in which a portion corresponding to a face region is indicated by a bounding box).

[0071] The liveness testing apparatus may test the liveness of an object based on a single input image. The single input image may correspond to a single picture, a single image, a still image of a single frame, etc. The liveness testing apparatus may test a liveness of an object included in a single input image by determining whether the object has one or more characteristics of a flat 2D surface or of a 3D structure. In more detail, for example, the liveness testing apparatus may test the liveness of the object included in the single input image by calculating a degree of uniformity in the distribution of light energy included in the object.

[0072] FIG. 3 illustrates a liveness testing apparatus 310 according to an example embodiment.

[0073] Referring to FIG. 3, the liveness testing apparatus 310 includes a receiver 311 and a tester 312.

[0074] In example operation, the receiver 311 receives an input image. The receiver 311 may receive an input image generated by an image sensor (not shown). The receiver 311 may be connected to the image sensor using a wire, wirelessly, or via a network. Alternatively, the receiver 311 may receive the input image from a storage device, such as a main memory, a cache memory, a hard disk drive (HDD), a solid state drive (SSD), a flash memory device, a network drive, etc.

[0075] The tester 312 tests a liveness of an object included in the input image. As discussed above, the tester 312 may test the liveness of the object by determining whether the object has one or more characteristics of a flat 2D surface or of a 3D structure. In one example, a successful test is one in which the object is determined to have one or more characteristics of a 3D structure, whereas a failed test is one in which the object is determined to have one or more characteristics of a flat 2D surface. In more detail, for example, the tester 312 may test the liveness of the object by analyzing a distribution of light energy included in the object of the input image. In a more specific example, the tester 312 may test the liveness of the object by calculating a degree of uniformity in the distribution of the light energy included in the object of the input image, and comparing the degree of uniformity of the distribution of light energy in the object of the input image with a threshold value. If the determined degree of uniformity is greater than or equal to a threshold value, then the object in the input image is determined to correspond to (have been obtained from) a flat 2D surface. On the other hand, if the determined degree of uniformity is less than or equal to the threshold value, then the object in the input image is determined to correspond to (have been obtained from) a 3D structure.

[0076] According to at least some example embodiments, the tester 312 may filter a plurality of pixels corresponding to the object included in the input image to analyze the distribution of the light energy included in the plurality of pixels corresponding to the object included in the input image. In one example, the tester 312 may filter the plurality of pixels using a diffusion process. In this example, the tester 312 may diffuse a plurality of pixels corresponding to the object included in the input image to analyze the distribution of the light energy included in the plurality of pixels corresponding to the object included in the input image. An example diffusion process will be described in more detail below with reference to FIG. 4.

[0077] Although example embodiments may be discussed in detail with regard to a diffusion process, it should be understood that any suitable filtering processes may be used in connection with example embodiments. In one example, example embodiments may utilize bilateral filtering. Because bilateral filtering is generally well-known, a detailed description is omitted. Moreover, any suitable filtering that preserves an edge region and blurs a non-edge region, in a manner similar or substantially similar to diffusion and bilateral filtering, may be used in connection with example embodiments discussed herein.

[0078] FIG. 4 illustrates a diffusion process according to an example embodiment.

[0079] According to at least some example embodiments, a liveness testing apparatus may diffuse a plurality of pixels corresponding to an object included in an input image. The liveness testing apparatus may iteratively update values of the plurality of pixels using a diffusion equation. In one example, the liveness testing apparatus may diffuse the plurality of pixels corresponding to the object included in the input image according to Equation 1 shown below.



[0080] In Equation 1, k denotes an iteration count, uk denotes a value of a pixel after a k-th iteration, and uk+1 denotes a value of a pixel after a (k+1)-th iteration. A value of a pixel in the input image is denoted u0.

[0081] Still referring to Equation 1, ∇ denotes a gradient operator, div() denotes a divergence function, and d() denotes a diffusivity function.

[0082] The diffusivity function d() may be given (or alternatively, desired or predetermined) function. In one example, the diffusivity function may be defined as shown below in Equation 2.



[0083] In Equation 2, β denotes a relatively small positive number (e.g., a minute value, such as about 10-6). When the diffusivity function defined as shown above in Equation 2 is used, a boundary of an object may be preserved relatively well during the diffusion process. When the diffusivity function is a function of pixel gradient ∇u as shown in Equation 2, the diffusion equation is a nonlinear diffusion equation.

[0084] The liveness testing apparatus may apply an additive operator splitting (AOS) scheme to solve Equation 1, and the liveness testing apparatus may diffuse the plurality of pixels corresponding to the object included in the input image according to Equation 3 shown below.



[0085] In Equation 3, I denotes a value of a pixel in an input image, Ax denotes a horizontal diffusion matrix, Ay denotes a vertical diffusion matrix, τ denotes a time step. A final iteration count L and the time step τ may be given (or alternatively desired or predetermined). In general, when the time step τ is set to be relatively small and the final iteration count L is set to be relatively large, a reliability of uL denoting a value of a finally diffused pixel may increase.

[0086] According to at least some example embodiments, the liveness testing apparatus may reduce the final iteration count L using the AOS scheme to solve Equation 1. When the AOS scheme is used, the reliability of uL, which is the value of the finally diffused pixel, may be sufficiently high although a time step τ of a given size is used. The liveness testing apparatus may increase efficiency of an operation for the diffusion process using the AOS scheme to solve the diffusion equation. The liveness testing apparatus may perform the diffusion process using a relatively small amount of processor and/or memory resources.

[0087] The liveness testing apparatus may effectively preserve a texture of the input image using the AOS scheme to solve the diffusion equation. The liveness testing apparatus may effectively preserve the original texture of the input image even in relatively low-luminance and backlit environments.

[0088] Referring to FIG. 4, image 410 corresponds to an input image, image 420 corresponds to an intermediate diffusion image, and image 430 corresponds to a final diffusion image. In this example, the final iteration count L is set to "20". The image 420 was acquired after values of pixels included in the input image were iteratively updated five times based on Equation 3. The image 430 was acquired after the values of the pixels included in the input image were iteratively updated 20 times based on Equation 3.

[0089] According to at least this example embodiment, the liveness testing apparatus may use a diffusion speed to determine whether the object included in the input image has one or more characteristics of a flat surface or of a 3D structure. The diffusion speed refers to a speed at which each pixel value is diffused. The diffusion speed may be defined as shown below in Equation 4.



[0090] In Equation 4, s(x, y) denotes a diffusion speed of a pixel at coordinates (x, y), u°(x, y) denotes a value of the pixel at coordinates (x, y) in an input image, and uL(x, y) denotes a value of the pixel at coordinates (x, y) in a final diffusion image. As shown in Equation 4, as a difference between a pixel value before diffusion and a pixel value after diffusion increases, a calculated diffusion speed also increase, whereas when the difference between the pixel value before diffusion and the pixel value after diffusion decreases, the calculated diffusion speed decreases.

[0091] More broadly, the liveness testing apparatus may determine whether the object included in the input image has one or more characteristics of a flat surface or of a 3D structure based on a magnitude of the change in pixel values in the image after L number of iterations of the above-discussed filtering process.

[0092] A face image may be divided into a small-scale region and a large-scale region. The small-scale region may refer to a region in which a feature point or a feature line is present. In one example, the small-scale region may include the eyes, the eyebrows, the nose, and the mouth of the face. The large-scale region may refer to a region in which a relatively large portion is occupied by skin of the face. In one example, the large-scale region may include the forehead and cheeks of the face.

[0093] Diffusion speeds of pixels belonging to the small-scale region may be greater than diffusion speeds of pixels belonging to the large-scale region. Referring back to the example shown in FIG. 4, a pixel 411 corresponding to eyeglass frames in the image 410 differs from neighboring pixels corresponding to skin, and thus, a value of the pixel 411 may change substantially (e.g., relatively greatly) as a result of diffusion. The value of the pixel 411 in the image 410 may be updated to a value of a pixel 431 in the image 430 by diffusion. Conversely, a pixel 412 corresponding to a cheek in the image 410 is similar to neighboring pixels, and thus, a value of the pixel 412 may change less than the value of the pixel 411(e.g., relatively slightly) by diffusion. The value of the pixel 412 in the image 410 may be updated to a value of a pixel 432 in the image 430 as a result of diffusion.

[0094] A difference in diffusion speeds may also result from a distribution of light energy in an image. When light energy in the image is more uniformly distributed, a relatively small diffusion speed may be calculated. Moreover, when the light energy is more uniformly distributed, the probability of neighboring pixels having similar pixel values may be higher (e.g., relatively high). Conversely, when light energy in the image is less (or non-) uniformly distributed, a relatively high diffusion speed may be observed. Moreover, when the light energy is less (or non-) uniformly distributed, the probability of neighboring pixels having different pixel values may be relatively high.

[0095] According to at least some example embodiments, a liveness testing apparatus may calculate a degree of uniformity in the distribution of the light energy in the image based on statistical information related to diffusion speeds. The liveness testing apparatus may test a liveness of an object in the image based on the statistical information related to diffusion speeds. To calculate the statistical information related to diffusion speeds, the liveness testing apparatus may extract a small-scale region from an image according to Equation 5 shown below.



[0096] In Equation 5, SR(x, y) is an indicator indicating whether a pixel at coordinates (x,y) belongs to a small-scale region. In this example, when a value of SR(x,y) corresponds to "1", the pixel at the coordinates (x,y) belongs to the small-scale region, whereas when the value of SR(x,y) corresponds to "0", the pixel at the coordinates (x,y) does not belong to the small-scale region.

[0097] The value of SR(x, y) may be determined based on a diffusion speed for the pixel at coordinates (x,y). For example, when the diffusion speed s(x,y) is greater than a given (or alternatively, desired or predetermined) threshold value, the value of SR(x,y) is determined to be "1". Otherwise, the value of SR(x,y) is determined to be "0". The threshold value may be set based on an average µ of the entire image and a standard deviation σ of the entire image. The average µ of the entire image may correspond to an average of diffusion speeds of pixels included in the entire image, and the standard deviation σ of the entire image may correspond to a standard deviation of the diffusion speeds of the pixels included in the entire image.

[0098] Hereinafter, an image in which a value of a pixel at coordinates (x,y) corresponds to SR(x,y) will be referred to as a small-scale region (SR) map. Since each pixel included in the SR map may have a value of "0" or "1", the SR map may also be referred to as a binary map. The SR map may effectively represent an underlying structure of a face in various illumination environments.

[0099] FIG. 5 illustrates two example SR maps according to example embodiments.

[0100] Referring to FIG. 5, as shown, the SR map 510 acquired when a medium displaying a face of a user is photographed differs from the SR map 520 acquired when the actual face of the same user is photographed. In the SR map 510 and the SR map 520, portions marked with black color correspond to pixels satisfying SR(x, y) = 1, whereas portions marked with white color correspond to pixels satisfying SR(x, y) = 0. In this example, in the SR map 510 and the SR map 520, the black portions have relatively fast diffusion speeds, and the white portions have relatively slow diffusion speeds.

[0101] According to at least one example embodiment, the liveness testing apparatus may test a liveness of a face in an image by analyzing an SR map. For example, the liveness testing apparatus may test the liveness of the face in the image by extracting various features from the SR map.

[0102] When an actual face of a user is photographed, various light reflections may occur due to curves on the actual face of the user.

[0103] When light energy in the image is less (or non-) uniformly distributed, a relatively large number of pixels having a pixel value of "1" may be included in the SR map. In this example, the liveness testing apparatus may determine the liveness of the face in the image based on Equation 6 shown below.



[0104] In Equation 6, N(SR(x, y) = 1) denotes a number of pixels satisfying SR(x, y) = 1, and ξ denotes a threshold value, which may be given, desired, or alternatively preset. The liveness testing apparatus may determine that the face in the image corresponds to a fake object when

is satisfied.

[0105] In another example, when light energy in the image is less (non-) uniformly distributed, a relatively large amount of noise components may be included in the SR map. In this case, the liveness testing apparatus may determine the liveness of the face in the image based on Equation 7 shown below.



[0106] In Equation 7, SRM(x, y) denotes a value of a pixel at coordinates (x, y) in an image acquired by applying median filtering to an SR map, and ξ denotes a threshold value, which may be given, desired, or alternatively preset. As the amount of noise components increases, a number of pixels having a difference between a value of SR(x, y) and a value of SRM(x, y) increase. In this example, when

is satisfied, the face in the image is determined to be a fake object.

[0107] Equations 6 and 7 are provided only as examples. The liveness testing apparatus may test the liveness of the object in the image based on a variety of diffusion speed based statistical information. In one example, the liveness testing apparatus may use a distribution of pixels having diffusion speeds greater than or equal to a given, desired or alternatively predetermined, threshold value.

[0108] In more detail, the liveness testing apparatus may determine the liveness of the face in the image based on Equation 8 shown below.



[0109] The liveness testing apparatus may determine that the face in the image corresponds to a live 3D object when ∑(x,y)N(SR(x,y) = 1) ≥ ξ is satisfied.

[0110] In another example, the liveness testing apparatus may determine the liveness of the face in the image based on Equation 9 shown below.



[0111] In this example, when ∑(x,y)|SR(x,y) - SRM(x,y)| ≥ ξ is satisfied, the face in the image is determined to be a live 3D object.

[0112] The liveness testing apparatus may also use diffusion speed based statistical information without using an SR map. In this example, the liveness testing apparatus may use respective values of diffusion speeds of all pixels, an average of the diffusion speeds of all the pixels, and a standard deviation of the diffusion speeds of all the pixels. The liveness testing apparatus may also use a filter response based on diffusion speeds. The liveness testing apparatus may use a result of applying median filtering to diffusion speeds of all pixels.

[0113] According to at least some example embodiments, the liveness testing apparatus may extract various features based on diffusion speed based statistical information, and learn the extracted features. The liveness testing apparatus may calculate diffusion speed based statistical information from various training images, and enable a classifier to learn the features extracted from the statistical information, in a learning stage. The training images may include images of a live 3D object and images of a fake 2D object.

[0114] A simply structured classifier may obtain a distance between vectors (e.g., a Euclidian distance) or a similarity (e.g., a normalized correlation), and compare the distance between the vectors or the similarity to a threshold value. A neural network, a Bayesian classifier, a support vector machine (SVM), or an adaptive boosting (AdaBoost) learning classifier may be used as a more elaborate classifier.

[0115] The liveness testing apparatus may calculate diffusion speed based statistical information from an input image, and extract features from the statistical information using a given, desired, or alternatively predetermined, method. The method may correspond to the method used in the learning stage. The liveness testing apparatus may input the extracted features and learned parameters into the classifier to test a liveness of the object included in the input image. Based on the extracted features and the learned parameters, the classifier may output a signal indicating whether the object included in the input image corresponds to a real object or a fake object.

[0116] FIG. 6 illustrates a liveness testing apparatus 600 according to an example embodiment.

[0117] Referring to FIG. 6, the liveness testing apparatus 600 includes: a receiver 611, a diffuser 612; and a tester 613. The receiver 611 may correspond to the receiver 311 shown in FIG. 3.

[0118] In example operation, the receiver 611 receives an input image, and outputs the input image to the diffuser 612 and the tester 613.

[0119] Although element 612 is referred to as a diffuser and the example embodiment shown in FIG. 6 will be described with regard to a diffusion operation, the element 612 may be more generally referred to as a filter or filter circuit 612. Moreover, any suitable filtering operation may be used as desired.

[0120] The diffuser 612 diffuses a plurality of pixels corresponding to an object included in the input image by iteratively updating values of the plurality of pixels corresponding to the object included in the input image based on a diffusion equation. In one example, the diffuser 612 may diffuse the plurality of pixels corresponding to the object included in the input image using Equation 1 discussed above.

[0121] The diffuser 612 may iteratively update the values of the plurality of pixels corresponding to the object included in the input image by applying an AOS scheme to the diffusion equation. In one example, the diffuser 612 may diffuse the plurality of pixels corresponding to the object included in the input image using Equation 3 discussed above. The diffuser 612 may output a diffusion image generated when the plurality of pixels are diffused.

[0122] Still referring to FIG. 6, the tester 613 tests a liveness of the object included in the input image based on diffusion speeds of the plurality of pixels. In one example, the tester 613 may test the liveness of the object by estimating a surface property related to the object based on the diffusion speeds. The surface property refers to a property related to a surface of the object, and may include, for example, a light-reflective property of the surface of the object, a number of dimensions of the surface of the object, and/or a material of the surface of the object.

[0123] The tester 613 may analyze a distribution of light energy included in the input image to estimate the surface property related to the object included in the input image. In one example, the tester 613 may analyze the distribution of the light energy included in the input image to determine whether the object included in the input image has a surface property (one or more characteristics) of a medium displaying a face (e.g., a 2D flat surface) or a surface property (one or more characteristics) of an actual face of a user (e.g., a 3D structure).

[0124] When the object included in the input image is determined to have a surface property of a medium displaying a face (e.g., a 2D flat surface), the tester 613 may output a signal corresponding to a failed test. That is, for example, when the tester 613 determines that the object included in the input image has a surface property of a medium displaying a face (e.g., a 2D flat surface), the tester 613 may output a signal indicative of a failed test. When the object included in the input image is determined to have a surface property of an actual face of a user (e.g., a 3D structure), the tester 613 may output a signal corresponding to a successful test. That is, for example, when the tester 613 determines that the object included in the input image has a surface property of an actual face of a user (e.g., a 3D structure), the tester 613 may output a signal indicative of a succesful test.

[0125] In another example, the tester 613 may test the liveness of the object by calculating statistical information related to the diffusion speeds. As discussed above, a 2D object and a 3D object have different light-reflective properties. And the different light-reflective properties between the 2D object and the 3D object may be modeled based on diffusion speeds.

[0126] In this example, the tester 613 may calculate a diffusion speed based on the input image from the receiver 611 and the diffusion image from the diffuser 612. In one example, the tester 613 may calculate a diffusion speed for each pixel using Equation 4 discussed above. To calculate statistical information related to diffusion speeds, the tester 613 may extract a small-scale region using Equation 5. The extracted small-scale region may be represented as an SR map. The tester 613 may then determine the liveness of the object included in the input image using Equation 6, 7, 8 or 9.

[0127] The tester 613 may test the liveness of the object included in the input image based on a variety of diffusion speed based statistical information. The tester 613 may use a distribution of pixels having diffusion speeds greater than or equal to a given, desired, or alternatively predetermined, threshold value. The tester 613 may also use diffusion speed based statistical information without using an SR map.

[0128] When the calculated statistical information corresponds to statistical information related to the medium displaying the face, the tester 613 may output a signal corresponding to a failed test. When the calculated statistical information corresponds to statistical information related to the actual face of the user, the tester 613 may output a signal corresponding to a successful test. That is, for example, when the calculated statistical information is indicative of a medium displaying the face, the tester 613 may output a signal indicating a failed test, whereas when the calculated statistical information is indicative of an actual face of the user, the tester 613 may output a signal indicating a successful test.

[0129] The liveness testing apparatus 600 may test the liveness of the object based on a single input image. The single input image may correspond to a single picture, a single image, or a still image of a single frame.

Image processing according to example embodiments



[0130] FIG. 7 is a flow diagram illustrating image processing according to example embodiments. FIG. 8 illustrates example changes in an input image depending on illumination according to example embodiments

[0131] Referring to FIG. 7, an input image 710 includes a face of a user. The face of the user included in the input image 710 is affected (e.g., greatly or substantially affected) by illumination. For example, referring to FIG. 8, although a face of the same user is photographed, different images may be generated depending on illumination. When an input image is vulnerable to changes in illumination, reliability of face recognition and/or user verification may decrease (e.g., considerably or substantially decrease), and/or a computational complexity may increase (e.g., substantially or considerably increase).

[0132] An image processing method and/or apparatus according to one or more example embodiments may generate an image less susceptible (e.g., impervious) to changes in illumination from an input image. An image processing method and/or apparatus according to one or more example embodiments may also generate an image that is substantially unaffected by illumination on an object when the input image is obtained. One or more example embodiments may provide technology that may generate an image less susceptible (e.g., impervious) to changes in illumination, thereby increasing reliability of face recognition and/or user verification, and/or reducing computational complexity of face recognition and/or user verification.

[0133] Still referring to FIG. 7, the input image 710 includes an illumination component 715 and a non-illumination component. In this example, the illumination component 715 refers to a component, from among components constituting pixel values, that is affected (e.g., substantially affected) by external illumination. The non-illumination component refers to a component, from among components constituting pixel values, that is substantially unaffected by external illumination. The image processing apparatus may separate the illumination component 715 from the input image 710 to generate an image less susceptible (e.g., impervious) to changes in illumination.

[0134] The image processing apparatus may detect a face region from an input image. In this example, example embodiments may be applicable to the face region detected from the input image. Hereinafter, the term "face image" refers to an input image including a face, or a face region extracted from the input image.

[0135] A face image may be expressed based on an illumination component and a non-illumination component. The face image may be based on a Lambertian model, as shown below in Equation 10.



[0136] In Equation 10, I denotes a face image, w denotes an illumination component, and v denotes a non-illumination component. With regard to the example shown in FIG. 7, I corresponds to the input image 710, w corresponds to the image 720 related to the illumination component, and v corresponds to the image 730 related to the non-illumination component.

[0137] The image 720 related to the illumination component may include the illumination component 715, whereas the image 730 related to the non-illumination component may not include the illumination component 715. Thus, the image 730 related to the non-illumination may be an image less susceptible (e.g., impervious) to changes in illumination. The image 730 related to the non-illumination component may also be referred to as a canonical image.

[0138] The illumination component 715 may have a relatively high probability of being distributed in a large-scale region of the image. Thus, the image 720 related to the illumination component may be an image corresponding to a large-scale region. The illumination component 715 may have a relatively low probability of being distributed in a small-scale region. Thus, the image 730 related to the non-illumination component may be an image corresponding to a small-scale region.

[0139] The image processing apparatus may generate the image 730 related to the non-illumination component based on the input image 710. In one example, the image processing apparatus may receive the input image 710, and generate the image 720 related to the illumination component based on the input image 710. The image processing apparatus may calculate the image 730 related to the non-illumination component based on the input image 710 and the image 720 related to the illumination component using Equation 10 shown above.

[0140] According to at least one example embodiment, the image processing apparatus may diffuse the input image 710 to generate the image 720 related to the illumination component. Diffusion speeds of pixels belonging to the small-scale region may be greater than diffusion speeds of pixels belonging to the large-scale region. The image processing apparatus may separate the small-scale region and the large-scale region based on a difference in diffusion speeds. The image processing apparatus may diffuse a plurality of pixels included in the input image 710 a number of times corresponding to a given, desired, or alternatively predetermined, iteration count (e.g., about 20) to generate the image 720 related to the illumination component corresponding to the large-scale region.

[0141] According to at least one example embodiment, the image processing apparatus may iteratively update values of the plurality of pixels using a diffusion equation. In one example, the image processing apparatus may diffuse the plurality of pixels corresponding to the face included in the input image 710 using Equation 11 shown below.



[0142] In Equation 11, k denotes an iteration count, uk denotes a value of a pixel after a k-th iteration, uk+1 denotes a value of a pixel after a (k+1)-th iteration, and uk corresponds to uk(x,y), which is a value of a pixel at coordinates (x,y) in an image after k diffusions. The value uk+l corresponds to uk+1(x,y), which is a value of a pixel at coordinates (x,y) in an image after (k + 1) diffusions. In this example, u0 denotes a value of a pixel in the input image 710. When a final iteration count corresponds to "L", uL denotes a value of a pixel in the image 720 related to the illumination component.

[0143] As before, ∇ denotes a gradient operator, div() denotes a divergence function, and d() denotes a diffusivity function. The diffusivity function may be given, desired, or alternatively predetermined. In one example, the image processing apparatus may define the diffusivity function as shown below in Equation 12.



[0144] In Equation 12, β denotes a small positive number. When the diffusivity function defined in Equation 12 is used, a boundary of a face may be preserved relatively well during the diffusion process. When the diffusivity function corresponds to a function of pixel gradient ∇u as shown in Equation 12, the diffusion equation is nonlinear. Herein, an image generated by diffusion is referred to as a diffusion image. When the diffusivity function is nonlinear, an image generated by diffusion is referred to as a nonlinear diffusion image.

[0145] Equation 12 is provided as an example of the diffusivity function, however, example embodiments may utilize other diffusivity functions. For example, one of a plurality of candidate diffusivity functions may be selected based on an input image.

[0146] Moreover, although example embodiments are discussed with regard to diffusivity functions, other filter functions may also be used, as mentioned above.

[0147] According to at least some example embodiments, the image processing apparatus may apply an AOS scheme to solve Equation 11. In one example, the image processing apparatus may diffuse the plurality of pixels corresponding to the face included in the input image 710 using Equation 13 shown below.



[0148] In Equation 13, I denotes a value of a pixel in the input image 710, Ax denotes a horizontal diffusion matrix, Ay denotes a vertical diffusion matrix, and τ denotes a time step. The final iteration count L and the time step τ may be given, desired, or alternatively predetermined. In general, when the time step τ is set to be relatively small and the final iteration count L is set to be relatively large, a reliability of uL, which denotes a value of a final diffused pixel, may increase.

[0149] Using the AOS scheme to solve Equation 11 may enable the image processing apparatus to reduce the final iteration count L. When the AOS scheme is used, the reliability of the final diffused pixel uL may be sufficiently high although the time step τ of a given, desired, or alternatively predetermined, size is used. Image processing apparatuses according to one or more example embodiments may increase an efficiency of operations for diffusion processes using the AOS scheme to solve diffusion equations.

[0150] The image processing apparatus may generate the image 730 related to the non-illumination component based on the input image 710 and the image 720 related to the illumination component. In one example, the image processing apparatus may generate the image 730 related to the non-illumination component using Equation 14 or 15. Since 'w' in Equation 10 corresponds to uL, Equations 14 and 15 may be derived from Equation 10.





[0151] In Equations 14 and 15, I denotes a face image, and may correspond to, for example, the input image 710. The face image I may also correspond to u0. The final diffused pixel uL denotes a large-scale region, and may correspond to, for example, the image 720 related to the illumination component. Still referring to Equations 14 and 15, v denotes a small-scale region, and may correspond to, for example, the image 730 related to the non-illumination component.

[0152] FIG. 9 illustrates an image processing apparatus 910 according to an example embodiment. Also shown in FIG. 9 is a face recognition and/or user verification circuit 920, which will be discussed in more detail later.

[0153] Referring to FIG. 9, the image processing apparatus 910 includes: a receiver 911; a diffuser 912; and a generator 913.

[0154] As with FIG. 6., although element 912 in FIG. 9 is referred to as a diffuser and the example embodiment shown in FIG. 9 will be described with regard to a diffusion operation, the element 912 may be more generally referred to as a filter or filter circuit 912. Moreover, as mentioned above, the filter 912 may utilize any suitable filtering operation, rather than the diffusion process discussed here.

[0155] In example operation, the receiver 911 may receive an input image. In a more specific example, the receiver 911 may receive an input image generated by an image sensor (not shown). The receiver 911 may be connected to the image sensor using a wire, wirelessly, or via a network. Alternatively, the receiver 911 may receive the input image from a storage device, such as, a main memory, a cache memory, a hard disk drive (HDD), a solid state drive (SSD), a flash memory device, a network drive, etc.

[0156] The diffuser 912 may diffuse a plurality of pixels corresponding to an object included in the input image. In one example, the object may correspond to a face of a user. The diffuser 912 may iteratively update values of the plurality of pixels corresponding to the object included in the input image based on a diffusion equation. In one example, the diffuser 912 may diffuse the plurality of pixels corresponding to the object included in the input image according to Equation 11.

[0157] The diffuser 912 may iteratively update the values of the plurality of pixels corresponding to the object included in the input image by applying an AOS scheme to the diffusion equation. In one example, diffuser 912 may diffuse the plurality of pixels corresponding to the object included in the input image using Equation 13. The diffuser 912 may output a diffusion image generated when the plurality of pixels are diffused. The diffusion image may correspond to an image related to an illumination component (e.g., 720 in FIG. 7).

[0158] Still referring to FIG. 9, the generator 913 may generate an output image based on the input image and the diffusion image. The generator 913 may generate the output image using Equation 14 or 15. The output image may correspond to an image related to a non-illumination component (e.g., 730 in FIG. 7). The generator 913 may output the output image to the face recognition and/or user verification circuit 920. The face recognition and/or user verification circuit 920 may perform any well-known face recognition and/or user verification circuit 920 operation as will be discussed in some detail later. Alternatively, the generator 913 may output the output image to a memory (not shown).

[0159] According to at least some example embodiments, the image processing apparatus 910 may generate the output image based on a single input image. The single input image may correspond to a single picture, a single image, or a still image of a single frame.

[0160] Still referring to FIG. 9, in one example the face recognition and/or user verification circuit 920 may recognize a face included in the input image based on the output image from the generator 913. The output image may correspond to an image related to the non-illumination component and less susceptible (e.g., impervious) to changes in illumination. The face recognition and/or user verification circuit 920 may recognize the face included in the input image based on an image less susceptible (e.g., impervious) to changes in illumination. Thus, accuracy and/or a reliability of the face recognition may increase. When an image less susceptible (e.g., impervious) to changes in illumination is used, a performance of an alignment operation may increase in a relatively low-luminance environment.

[0161] In another example, the face recognition and/or user verification circuit 920 may verify the user based on the output image from the generator 913. The image processing apparatus 910 may verify the user by recognizing the face of the user based on the output image. The output image may correspond to an image related to the non-illumination component and less susceptible (e.g., impervious) to changes in illumination. The image processing apparatus 910 may verify the user based on an image less susceptible (e.g., impervious) to a change in illumination. Thus, accuracy and/or a reliability of the user verification may increase.

Flowchart according to example embodiments



[0162] FIG. 10 is a flow chart illustrating a liveness testing method according to an example embodiment. In some cases, the flow chart shown in FIG. 10 will be discussed with regard to the liveness testing apparatus shown in FIGS. 3 and 6.

[0163] Referring to FIG. 10, at operation 1010 the liveness testing apparatus receives an input image. At operation 1020, the liveness testing apparatus tests a liveness of an object included in the received input image.

[0164] With regard to the liveness testing apparatus 310 shown in FIG. 3, for example, at operation 1010, the receiver 311 receives the input image, and the tester 312 tests a liveness of the object included in the received input image at operation 1020. The tester 312 may test the liveness of the object included in the input image received at the receiver 311 based on whether the object in the input image has one or more characteristics of a flat surface or a 3D structure. The details of the operation performed by the tester 312 are discussed above with regard to FIG. 3, and thus, a detailed discussion is not repeated here.

[0165] With regard to the liveness testing apparatus 600 shown in FIG. 6, for example, at operation 1010 the receiver 611 receives the input image. At operation 1020, the diffuser 612 diffuses a plurality of pixels corresponding to the object included in the input image received at the receiver 611, and the tester 613 tests the liveness of the object based on diffusion speeds of the plurality of pixels. The details of the operation performed by the diffuser 612 and the tester 613 are discussed above with regard to FIG. 6, and thus, a detailed discussion is not repeated here.

[0166] FIG. 11 is a flow chart illustrating an image processing method according to an example embodiment. For example purposes, the image processing method shown in FIG. 11 will be discussed with regard to the image processing apparatus shown in FIG. 9.

[0167] Referring to FIG. 11, at operation 1110, the image processing apparatus receives a first image. At operation 1120, the image processing apparatus generates a second image, and at operation 1130 the image processing apparatus generates a third image.

[0168] The first image may correspond to an input image (e.g., 710 in FIG. 7), the second image may correspond to an image related to an illumination component (e.g., 720 in FIG. 7), and the third image may correspond to an image related to a non-illumination component (e.g., 730 in FIG. 7).

[0169] In more detail, with regard to FIGS. 9 and 11, for example, at operation 1110 the receiver 911 receives a first (input) image.

[0170] At operation 1120, the diffuser 912 generates a second image based on the first (input) image. In this example, the second image is an image related to an illumination component (e.g., 720 in FIG. 7).

[0171] At operation 1130, the generator generates a third (output) image based on the first (input) image and the second image generated by the diffuser 912. In this case, the third (output) image is an image related to a non-illumination component (e.g., 730 in FIG. 7). The details of the operations performed by the diffuser 912 and the generator 913 are discussed above with regard to FIG. 9, and thus, a detailed discussion is not repeated here.

[0172] More generally, the descriptions provided with reference to FIGS. 1A through 9 may be applicable to operations of FIGS. 10 and 11, and thus, more detailed descriptions are omitted for conciseness.

[0173] FIG. 12 illustrates an image processing method according to another example embodiment.

[0174] The example embodiment shown in FIG. 12 combines the liveness testing method of FIG. 10 with the image processing method of FIG. 11. For example purposes, the method shown in FIG. 12 will be described with regard to the image processing apparatus shown in FIG. 9. The details of the operations described with regard to FIG. 12 are provided above with regard to, for example, FIGS. 3, 6, 9, 10 and 11, and thus, a detailed discussion is not repeated here.

[0175] Referring to FIG. 12, at operation 1210 the receiver 911 of the image processing apparatus 910 receives a first image. The first image may correspond to an input image including a face of a user. The receiver 911 outputs the first image to the diffuser 912 and the generator 913.

[0176] At operation 1220, the diffuser 912 generates a second image based on the first image received at the receiver 911. The diffuser 912 generates the second image by diffusing the first image from the receiver 911. The second image may be an image related to an illumination component.

[0177] At operation 1240, the generator 913 calculates a diffusion speed for each pixel based on the first image and the second image. The diffusion speed for each pixel may be calculated based on a difference between a pixel value in the second image and a corresponding pixel value in the first image.

[0178] At operation 1250, the generator 913 extracts statistical information based on diffusion speeds. For example, the generator 913 calculates a number of pixels having diffusion speeds greater than a given, desired, or alternatively predetermined, threshold value.

[0179] At operation 1270, the generator 913 performs a liveness test based on the diffusion speed based statistical information. In one example, the generator 913 determines whether the input image corresponds to a real 3D object based on the number of pixels having diffusion speeds greater than the threshold value.

[0180] If the generator 913 determines that the input image does not correspond to a real 3D object (the liveness test fails), then the face recognition and/or user verification circuit 920 does not perform face recognition and/or user verification at operation 1260, and the process terminates.

[0181] Returning to operation 1270, if the generator 913 determines that the input image does correspond to a live 3D object (the liveness test succeeds), then face recognition and/or user verification is performed. In this example, an image less susceptible (e.g., impervious) to changes in illumination may be generated for use in face recognition and/or user verification operations.

[0182] Still referring to FIG. 12, at operation 1230 the generator 913 generates a third image based on the first image and the second image. In one example, the generator 913 calculates the third image based on a ratio of the first image to the second image as discussed above with regard to Equation 14 or a difference between the first image and the second image in a log domain as discussed above with regard to Equation 15. The third image may be an image related to a non-illumination component and impervious to a change in illumination.

[0183] At operation 1260, the face recognition and/or user verification circuit 920 performs face recognition and/or user verification based on the third image. In the example embodiment shown in FIG. 12, the face recognition and/or user verification circuit 920 performs the face recognition and/or user verification operations in operation 1260 only when the input image does correspond to a real 3D object (the liveness test succeeds). In this example, the face recognition and/or user verification may be performed based on the third image corresponding to the image less susceptible (e.g., impervious) to changes in illumination.

[0184] Details of the descriptions provided with reference to FIGS. 1A through 11 may be applicable to operations of FIG. 12, and thus, duplicated descriptions are omitted for conciseness.

[0185] FIG. 13 is a block diagram illustrating an electronic system according to an example embodiment.

[0186] Referring to FIG. 13, the electronic system includes, for example: an image sensor 1300, an image signal processor (ISP) 1302, a display 1304 and a memory 1308. The image sensor 1300, the ISP 1302, the display 1304 and the memory 1308 communicate with one another via a bus 1306.

[0187] The image sensor 1300 may be the image sensor 115 described above with regard to FIGS. 1A and 1B. The image sensor 1300 is configured to capture an image (also referred to as image data) in any well-known manner (e.g., by converting optical images into electrical signals). The image is output to the ISP 1302.

[0188] The ISP 1302 may include one or more of the apparatuses and/or may perform one more of the methods discussed above with regard to discussed above with regard to FIGS. 1A through 12. The ISP 1302 may also include the face recognition and/or user verification circuit 920 to perform face recognition and/or user verification operations discussed above with regard to FIGS. 1A through 12. In a more specific example, the ISP 1302 may include the liveness testing apparatus 310 shown in FIG. 3, the liveness testing apparatus 610 shown in FIG. 6, the image processing apparatus 910 shown in FIG. 9 and/or the face recognition and/or user verification circuit 920 shown in FIG. 9. The memory 1308 may store images captured by the image sensor 1300 and/or generated by the liveness testing apparatus and/or the image processing apparatus. The memory 1308 may be any suitable volatile or non-volatile memory. The display 1304 may display images captured by the image sensor 1300 and/or generated by the liveness testing apparatus and/or the image processing apparatus.

[0189] The ISP 1302 may also be configured to execute a program and control the electronic system. The program code to be executed by the ISP 1302 may be stored in the memory 1308.

[0190] The electronic system shown in FIG. 13 may be connected to an external device (e.g., a personal computer or a network) through an input/output device (not shown) and may exchange data with the external device.

[0191] The electronic system shown in FIG. 13 may embody various electronic systems including: a mobile device, such as a mobile phone, a smartphone, a personal digital assistant (PDA), a tablet computer, a laptop computer, etc.; a computing device, such as a personal computer (PC), a tablet PC, a netbook; or an electronic product, such as a television (TV) or smart TV, a security device for a gate control, etc.

[0192] One or more example embodiments (e.g., liveness testing apparatuses, image processing apparatuses electronic systems, etc.) described herein may be implemented using hardware components and software components. For example, the hardware components may include microphones, amplifiers, band-pass filters, audio to digital convertors, and processing devices. A processing device may be implemented using one or more special purpose computers, such as, for example, a processor, a controller and an arithmetic logic unit, application-specific-integrated-circuit, system-on-chip device, a digital signal processor, a microcomputer, a field programmable array, a programmable logic unit, a microprocessor or any other device capable of responding to and executing instructions in a defined manner. The processing device may run an operating system (OS) and one or more software applications that run on the OS. The processing device also may access, store, manipulate, process, and create data in response to execution of the software. For purpose of simplicity, the description of a processing device is used as singular; however, one skilled in the art will appreciated that a processing device may include multiple processing elements and multiple types of processing elements. For example, a processing device may include multiple processors or a processor and a controller. In addition, different processing configurations are possible, such a parallel processors.

[0193] Furthermore, example embodiments may be implemented by hardware, software, firmware, middleware, microcode, hardware description languages, or any combination thereof. When implemented in software, firmware, middleware or microcode, the program code or code segments to perform the necessary tasks may be stored in a machine or computer readable medium such as a computer readable storage medium. When implemented in software, a processor or processors will perform the necessary tasks.

[0194] A code segment may represent a procedure, function, subprogram, program, routine, subroutine, module, software package, class, or any combination of instructions, data structures or program statements. A code segment may be coupled to another code segment or a hardware circuit by passing and/or receiving information, data, arguments, parameters or memory contents. Information, arguments, parameters, data, etc. may be passed, forwarded, or transmitted via any suitable means including memory sharing, message passing, token passing, network transmission, etc.

[0195] The software may include a computer program, a piece of code, an instruction, or some combination thereof, to independently or collectively instruct or configure the processing device to operate as desired. Software and data may be embodied permanently or temporarily in any type of machine, component, physical or virtual equipment, computer storage medium or device, or in a propagated signal wave capable of providing instructions or data to or being interpreted by the processing device. The software also may be distributed over network coupled computer systems so that the software is stored and executed in a distributed fashion. The software and data may be stored by one or more non-transitory computer readable recording mediums.

[0196] Example embodiments described herein may be recorded in non-transitory computer-readable media including program instructions to implement various operations embodied by a computer. The media may also include, alone or in combination with the program instructions, data files, data structures, and the like. The program instructions recorded on the media may be those specially designed and constructed for the purposes embodied herein, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of non-transitory computer-readable media include magnetic media such as hard disks, floppy disks, and magnetic tape; optical media such as CD ROM discs and DVDs; magneto-optical media such as optical discs; and hardware devices that are specially configured to store and perform program instructions, such as read-only memory (ROM), random access memory (RAM), flash memory, and the like. Examples of program instructions include both machine code, such as produced by a compiler, and files containing higher level code that may be executed by the computer using an interpreter. The above-described devices may be configured to act as one or more software modules in order to perform the operations of the above-described example embodiments, or vice versa.

[0197] A number of examples have been described above. Nevertheless, it should be understood that various modifications may be made. For example, suitable results may be achieved if the described techniques are performed in a different order and/or if components in a described system, architecture, device, or circuit are combined in a different manner and/or replaced or supplemented by other components or their equivalents. Accordingly, other implementations are within the scope of the following claims.


Claims

1. An image processing method comprising:

receiving a first image including an illumination component and a non-illumination component;

filtering a plurality of pixels included in the first image to generate a second image related to the illumination component; and

generating a third image related to the non-illumination component based on the first image and the second image.


 
2. The method of claim 1, wherein the second image is a diffusion image, and the filtering comprises:
diffusing the plurality of pixels included in the first image to generate the diffusion image.
 
3. The method of claim 2, wherein the diffusing comprises:
iteratively updating values of the plurality of pixels by applying an additive operator splitting (AOS) scheme to a diffusion equation.
 
4. The method of claim 2 or 3, wherein a diffusion speed of a pixel corresponding to the non-illumination component is higher than a diffusion speed of a pixel corresponding to the illumination component.
 
5. The method of any one of the previous claims, further comprising at least one of:

recognizing a face based on the third image; and

verifying a user based on the third image.


 
6. The method of any one of the previous claims, wherein the third image is generated based on at least one of a ratio of the first image to the second image and a difference between the first image and the second image in a log domain.
 
7. The method of any one of the previous claims, wherein the non-illumination component is included in a first scale region impervious to a change in illumination, and the illumination component is included in a second scale region sensitive to a change in illumination.
 
8. The method of claim 7, wherein the first scale region is a small-scale region of the first image, and wherein the second scale region is a large-scale region of the first image.
 
9. The method of claim 8, wherein the small-scale region comprises a feature point or a feature line, preferably including any one of eyes, eyebrows, nose and mouth of the face, and wherein the large-scale region comprises a relatively large portion of skin, preferably including any one of forehead and cheeks of the face.
 
10. The method of any one of the previous claims, wherein the first image corresponds to a single image of a face.
 
11. A non-transitory computer-readable medium comprising a program that, when executed on a computer device, causes the computer device to perform the method of any one of the previous claims.
 
12. An image processing apparatus comprising:

a receiver configured to receive a first image including an illumination component and a non-illumination component;

a filter circuit configured to filter a plurality of pixels included in the first image to generate a second image related to the illumination component; and

a generator circuit configured to generate a third image related to the non-illumination component based on the first image and the second image.


 
13. The apparatus of claim 12, wherein the first image corresponds to a single image including a face.
 
14. The apparatus of claim 12 or 13, wherein the second image is a diffusion image, and the filter circuit comprises:
a diffuser configured to diffuse the plurality of pixels included in the first image to generate the diffusion image.
 
15. The apparatus of claims 12, 13 or 14, wherein the diffuser is configured to iteratively update values of the plurality of pixels corresponding to the object included in the first image based on a diffusion equation.
 




Drawing