METHOD FOR DETECTING THE QUALITY OF THE MULTIMEDIA COMMUNICATION

Description

Field of the Invention

[0001] The present invention relates to multimedia communication technology, and particularly to a method for measuring multimedia video communication quality.

Background of the Invention

[0002] With the arrival of multimedia information age, various multimedia processing and communication technologies, especially video processing and video communication technologies emerge. Accordingly, quality evaluation for digital video is becoming more and more important. The digital video is the most primary media form for multimedia video communication nowadays. A network streaming media (e.g. a network movie, a network TV and so on), a video conference and a video telephone etc. are all multimedia applications based on video. Customer satisfaction on a multimedia application service relies on video quality to a great extent. Digital video signal transmission includes some processes such as sampling, quantifying, compression coding, network transmitting, decoding and restoring for an analog video signal, wherein in some processes, especially, network transmitting without guaranteeing QoS(Quality of Service) (e.g. packet transmission network), errors and information distortion may be introduced in each process, thus leading to a decreased customer satisfaction. Multimedia video communication quality is a measurement for measuring distortion of a digital multimedia signal relative to an original signal. Video quality evaluation plays a very important role in the field of video compression, processing and video communication. Performance of a real-time or non real-time video system and the QoS of various video communication transmission channels are finally reflected by the video quality, and feedback for adjusting parameters of a codec or a channel is given, thus guaranteeing the video quality within an acceptable range. An easy to understand measurement for output video quality of various different codecs is presented so that the performance of the codec may be designed, evaluated and optimized. A graphics and image display system according with a human visual model is designed and optimized.

[0003] Two factors affecting the video quality in a wireless and IP (Internet Protocol) transmission video system are: one is that the video quality is degraded due to video compression; the other is that the video quality is affected by channel packet loss and random error codes due to relatively bad channel environment, especially video contents are damaged by various errors of header information and motion vectors. Some unendurable mosaic blocks are often formed in these damaged video contents, and greatly affect the subjective quality of the recovered video. Therefore, video quality measurement, feedback correction and test development based on the measurement are desired for the current multimedia video communication evolving towards a direction of wireless and IP transmission.

[0004] The video quality evaluation is also very significant for video communication equipment manufacturers and telecom operators. For the equipment manufacturers, providing a convictive video quality evaluation result of a system may greatly facilitate the sale of their products. For the operators, the evaluation data of the video quality may be used for popularization and propagation of their services. In addition, developing an automatic real-time video quality evaluation method, based on which both the manufacturers and the operators are capable of performing real-time monitoring to video equipment, will facilitate trouble shooting and fault diagnosis.

[0005] The video quality evaluation may be classified into subjective quality evaluation and objective quality evaluation. A result of the subjective quality evaluation is reliable, but the evaluation for video subjective quality is strict with a Human Test Subject because it relies on participation of the Human Test Subject, the process of which is complicated and hard to be widely applied, especially for evaluation in an application environment with real-time demand. The objective quality evaluation measures the video quality using a quantifying method from another aspect, and may be performed automatically with high efficiency and without human participation.

[0006] The objective quality evaluation may be further classified into three categories:

[0007] (1) Full Reference Model in need of a full original video sequence;

[0008] (2) Partial Reference Model only in need of partial statistic characteristics of an original video sequence;

[0009] (3) Referenceless Model without need of any information of an original video sequence.

[0010] The full reference and the partial reference evaluation methods are not widely used because the needed reference video sequence generally cannot be obtained in practical applications, so a new objective real-time video quality evaluation method without need of any reference information is desired. At present, the VQEG (Video Quality Experts Group) is dedicated to research methods and constitute standards for the referenceless video quality evaluation.

[0011] PSNR(Peak Signal-to-Noise Ratio) is most widely used among numerous objective quality evaluation indications, because it is easy to be calculated, has apparent physical meaning and may actually reflect a distortion degree of an image.

[0012] Digital media is easy to be accessed, copied, transmitted and edited, while some problems such as pirating a digital media copyright and interpolating digital media contents will arise. Thus a digital watermark technique early used for protecting a digital media copyright is presented.

[0013] The digital watermark technique is developed and widely used in recent years. Watermark information embedded in original media data always coexists with the original media data by embedding a series of information in the original media data, thus the copyright of the original media data and the integrity of contents may be protected. With development of technologies, besides copyright protection, the digital watermark technique may be used in many other fields. For example, the multimedia video communication quality measurement mentioned above may be achieved by embedding and extracting a digital watermark, which is a Partial Reference Model method, without knowing an original video image but indirectly reflecting status of the original image with reference to a watermark image before channel transmission and a distorted watermark image after channel transmission, accordingly the quality measurement, calculation and evaluation are performed.

[0014] The digital watermark may be embedded in any portion of the original data, but its effect on the original data needs to be reduced to the utmost extent. The digital watermark technology may be classified into a digital watermark technology of the spatial domain and a digital watermark technology of the transform domain according to digital watermark embedding approaches. In the digital watermark technology, watermark information is directly embedded in the spatial domain of a media. For example, the information is embedded in pixels of an image. In the digital watermark technology of the transform domain, a transform is performed on a media firstly, such as DFT (Discrete Fourier Transform), DCT (Discrete Cosine Transform) or DWT (Discrete Wavelet Transform) etc., and then the watermark information is embedded in the transform domain.

[0015] FIG. 1 shows a block diagram illustrating a principle of a digital watermark. In this figure, the master media I₀ generally is original or compressed multimedia data such as video or audio etc., and the data b₀ to be hidden only has less data in comparison with I₀. The difference between the media I₁ embedded with a watermark and I₀ is distortion caused by the embedded watermark. In general, such distortion is not desired to be perceived by human. The media I₂ is obtained by performing some processing on I₁, such as data compression, noise contamination and intended attacks to the watermark, which may be regarded as noise. Therefore, the watermark b₁ extracted from I₂ may be distorted to a certain extent in comparison with the original watermark b₀ If I₂ is identical with I₁, the watermark b₁ extracted from I₂ should be identical with the original watermark b₀.

[0016] A common mathematical model for watermark embedding and extracting is: I₀ and I₁ respectively represent the original data and the data embedded with the watermark, b₀ represents the original watermark, the watermark embedding process may be expressed as: I₁ = I₀ + f (I₀, b₀), wherein f (I₀, b₀) represents a watermark embedding algorithm. The watermark measurement process may be expressed as: if H₀ : b₁ = I₂ -I₀ = N , the watermark does not exist; if H₁ : b₁ = I₂ - I₀ = b₀ + N, the watermark exists. N represents noise. For example, the noise may be caused by data compression, noise contamination and intended attacks to the watermark etc. The data embedded with the watermark will be distorted to a certain extent after processing, so the watermark measured from the processed data may be different from the original watermark to a certain extent.

[0017] Watermark measurement technology is commonly implemented by the classical Signal Detection technology, which is used for researching how to determine whether a destination signal exists in noise, such as whether a reflecting signal from a destination is contained in a radar echo signal, and if the destination signal exists, how to perform optimal signal extraction using a statistic principle. Statistic Hypothesis Test/Validation is used for determining whether a signal exists in noise. A watermark measurement process includes: presenting two hypotheses H₀ and H₁, determining which hypothesis is true according to a test result, and accordingly knowing whether a watermark exists.

[0018] At present, all the methods for multimedia (video and images) quality evaluation based on digital watermark comply with a same principle. FIG.2 shows a block diagram illustrating a principle of a video communication quality measurement based on a digital watermark, and the basic principle of which will be described hereinafter.

[0019] A watermark image, which is very small relative to an original image, is embedded in the original image, thus the added additional data quantity is negligible in comparison with the data quantity of the original image, and has little effect on the multimedia quality. The original image embedded with the watermark is distorted after passing through a transmission channel, and the watermark image extracted from which will be accordingly distorted. The watermark is embedded in the original image everywhere, so the quality measurement for the watermark image is equivalent to the measurement for the original image after distributed sampling. As long as the distribution of the watermark is uniform enough, the quality of the distorted image may be sufficiently reflected by measuring the quality of the watermark image. While the original image is shared by both sides of a communication, the receiving side measures communication quality by comparing the recovered watermark image with the known original watermark image, thus the quality evaluation for the watermark is actually a kind of evaluation with reference.

[0020] FIG.3 shows an example of measuring the multimedia video communication quality based on a digital watermark. The left is an original watermark image, which is a black-and-white (easy to determine error positions) bi-level image. The middle is a distorted image of an international standard test image Lena after being embedded with the watermark and JPEG (Joint Photographic Experts Group) standard compression and network packet loss during a transmission process, wherein there are three obvious block distortions in the image. The right is a recovered watermark extracted from the distorted Lena image, wherein scatteredly distributed black spots and white spots are distortions due to the compression, while big black and white blocks, the positions of which correspond to three obvious distortions in the recovered Lena image, are caused by errors such as packet loss.

[0021] In practical applications, the watermark image embedment inevitably results in a degradation of the validity of the multimedia video data and the video quality when the above-mentioned solution is used. Theoretically, the more watermark information are embedded in video, the better the whole image may be covered by the watermark information, and the more samples may be presented. Accordingly the extracted watermark information may fully reflect the original video, and the obtained video quality evaluation is more accurate. On the other side, the more watermark information is embedded in video, the more the quality of the video is affected. Thus the video quality is decreased. Namely, there is a contradiction in the video quality evaluation methods based on a digital watermark: the more information is embedded in video, the more accurate the video quality measurement is, but the more the embedded watermark information affects the video quality. The prior art is unable to solve the contradiction, so the video communication quality may not be measured and calculated accurately by embedding and extracting a watermark under a precondition that the video communication quality is not affected.

[0022] The main reason for this instance is that the watermark information is not embedded uniformly in the multimedia video data everywhere in the prior art, which leads to the multimedia video communication quality is not fully reflected by the small quantity of watermark information, while too much watermark information will affect the multimedia video data.

[0023] In CAMPISI P ET AL: "BLIND QUALITY ASSESSMENT SYSTEM FOR MULTIMEDIA COMMUNICATIONS USING TRACING WATERMARKING" IEEE transactions on signal processing, IEEE SERVICE CENTER, NEW YORK, NY, US, vol.51, no.4, April 2003 (2003-04), pages 996-1002, it is disclosed a method to blindly estimate the quality of a multimedia communication link by means of an unconventional use of digital fragile watermarking. In this method, a fragile watermark is hidden in an MPEG-like host data video transport stream using a spread-spectrum approach.

[0024] U.S. patent application US 2002/157005 A1 discloses a method for authenticating a media signal and related software, systems and applications. A digital watermark is embedded in the media signal. A metric is calculated for the digital watermark and then embedded in the media signal with the digital watermark. To detect a potential alteration, a detector computes the metric for a potentially corrupted version of the embedded media signal, and then compares its computed metric to the embedded metric to detect whether the alteration has occurred.

[0025] U.S. patent application US 2003/112996 A1 discloses a method for dynamic monitoring and correcting network or transcoding service quality by use of watermarking techniques.

Summary of the Invention

[0026] According to an embodiment of the present invention, there is provided a method for measuring multimedia video communication quality. The method is an objective quality evaluation, so that the multimedia video communication quality may be objectively reflected by embedding and extracting a digital watermark under a precondition that the quality of the multimedia video data is not obviously affected

[0027] According to an embodiment of the present invention, there is provided a method for measuring multimedia video communication quality, including:

[0028] embedding original watermark data in multimedia video data uniformly at a transmitting side, and transmitting the multimedia video data to a receiving side;

[0029] extracting a recovered watermark from the received multimedia video data at the receiving side;

[0030] calculating a distortion degree of the recovered watermark relative to the original watermark.

[0031] Preferably, the method may also includes: uniformly embedding the original watermark in the original multimedia data, performing compression coding or data packaging or framing at the transmitting side, and transmitting the data to the receiving side.

[0032] Preferably, the method may also include: uniformly dividing the multimedia video data frames into blocks of equal size, and embedding equal amount of the original watermark data in a part of the blocks or all of the blocks at the transmitting side.

[0033] The method further includes: dividing the multimedia video data into groups at the transmitting side, each group including equal number of frames, and selecting at least two frames with equal interval in a same group as embedded frames, uniformly dividing each of the embedded frame into a plurality of blocks of equal size, selecting equal number of blocks from the plurality of blocks as embedded blocks, and embedding equal amount of the original watermark data in each of the embedded block.

[0034] Each multimedia video data frame includes M rows × N columns pixels in which the M rows are numbered from 0 to M-1 and the N columns are numbered from 0 to N-1, and uniformly dividing a multimedia video data frame into blocks of equal size includes:

[0035] dividing each multimedia video data frame into m strips numbered from 0 to m-1, the strip i (i = 0,1,L , m - 2) including pixels from row i × k_v to row (i + 1) × k_v - 1, and the strip m-1 including pixels from row (m-1)×k_v to row M - 1, wherein m is a minimum integer greater than or equal to M/k_v ; if m = M/k_v, the strip m-1 including zero row of pixels;

[0036] dividing the strip i(i = 0,1,L , m -1) into n blocks numbered from i × n to (i + 1)×n-1, the block i × n + j (0,1,L , n - 2) including pixels from column j × k_h to column (j + 1)×k_h -1 of this strip, and the block n-1 including pixels from column (n -1)× k_h to column N-1, wherein n is a minimum integer greater than or equal to N/k_h ; if n = N/k_h , the strip n-1 including zero column of pixels;

[0037] embedding equal amount of the original watermark data in each of the obtained m × n blocks.

[0038] Each frame of the multimedia video data is uniformly divided into L blocks numbered from 0 to L-1, and the embedding the original watermark further includes:

[0039] dividing a frame sequence into groups, each group including P × Q frames, which sequentially are frame 0 to frame P × Q-1 of the group;

[0040] selecting a frame q × P (q = 0,L , Q-1) of each group as the embedded frame;

[0041] selecting a block q + r × Q (r = 0,L , R -1, and q + r × Q ≤ L) of the embedded frame q × P (q = 0,L , Q-1) as the embedded block, wherein R is a maximum integer less than or equal to L/Q;

[0042] embedding equal amount of the original watermark data in all the embedded blocks of the group.

[0043] Preferably, the method may also includes: determining a bit number of the original watermark data embedded in a part of the blocks or all of the blocks according to the size of the data block.

[0044] Embedding 1-bit data of the original watermark in the multimedia video data block includes:

[0045] calculating an average value µ of the multimedia video data block according to pixels;

[0046] modifying a value of each pixel in the multimedia video data block according to b' = b+c, wherein, b' is a modified value of a pixel, b is a value of the pixel before modification, and c is a modification amount;

[0047] the modification amount is calculated according to the following formula:

[0048] when the original watermark data bit is 1,

[0049] when the original watermark data bit is 0,

[0050] wherein γ is a remainder of µ being divided by 2A , A represents an embedding strength and is equal to 2 × Q_P -1, and Q_P represents a quantization factor for the multimedia video communication compression coding.

[0051] Extracting a data bit of the recovered watermark from the received multimedia video data block includes:

[0052] calculating an average value µ' of the received multimedia video data block according to pixels;

[0053] calculating extracted data w' of the recovered watermark according to the following formula:

[0054] wherein γ' is a remainder of µ' being divided by 2A , A represents an embedding strength and is equal to 2 × Q_P -1, and Q_P represents a quantization factor for the multimedia video communication compression coding.

[0055] The method further includes:

[0056] calculating a Peak Signal-to-Noise Ratio using a Pixel Recovery Rate of the recovered watermark relative to the original watermark to measure objective quality of the multimedia video communication;

[0057] fitting to obtain an estimated value of the Peak Signal-to-Noise Ratio using a linear function of the Pixel Recovery Rate of the recovered watermark relative to the original watermark based on statistic data of a multimedia video communication experiment and according to a Mean Square Error criterion to measure the objective quality of the multimedia video communication.

[0058] Preferably, the method may also include: calculating the distortion degree of the recovered watermark relative to the original watermark to measure the objective quality of the multimedia video communication.

[0059] The method also includes: calculating the distortion degree of the recovered watermark relative to the original watermark using the Pixel Recovery Rate of the recovered watermark relative to the original watermark.

[0060] Preferably, the method may also include: on extracting the recovered watermark, post-processing the received multimedia video data to eliminate effect of the watermark embedment, wherein the post-processing includes one of the following

[0061] smoothing filtering, median filtering, neural network filtering, mathematical morphology filtering and filtering based on fuzzy mathematics methods.

[0062] The method also includes:

[0063] determining the original watermark, relevant parameters and a strategy for embedding the original watermark and extracting the recovered watermark through a communication negotiation between the transmitting side and the receiving side.

[0064] determining the original watermark, the relevant parameters and the strategy for embedding the original watermark and extracting the recovered watermark at the transmitting side, and informing the receiving side before communication.

[0065] For a color image, the original watermark is embedded in a luminance component of the original multimedia data.

[0066] For a multi-channel image, the original watermark is embedded in one or more channels of the original multi-channel image data.

[0067] The original watermark is a bi-level image.

[0068] Compared with the prior art, according to the technical solution of the embodiments of the present invention, each frame of the multimedia video data is uniformly divided into blocks of equal size, and the watermark data is embedded in each block for completely uniform watermark distribution. Thus the accuracy of reflecting the multimedia video communication quality by the watermark is improved.

[0069] Moreover, frames of the multimedia video data are divided into groups and the watermark is uniformly embedded with equal interval in a part of the frames in each group for completely uniform watermark distribution. The amount of the watermark information is decreased. The effect of the watermark embedment on the multimedia video data is reduced. The multimedia video communication quality is ensured and measured accurately.

[0070] In addition, the watermark information is directly and uniformly embedded in the original multimedia data or the video spatial domain, so that the watermark may be completely equivalent to an intuitive quality reflection of multimedia or video, and the accuracy of the multimedia video communication quality evaluation is improved.

[0071] Post-processing such as smoothing filtering is performed on the video image after watermark extraction to eliminate the effect of the watermark on the multimedia video data quality.

[0072] A linearly fitted PSNR of the multimedia video data using the accuracy rate of the recovered watermark relative to the original watermark is regarded as a measurement standard for the multimedia video communication quality, which is able to reflect the communication status accurately.

[0073] Post-processing such as smoothing filtering, median filtering, neural network filtering, mathematical morphology filtering or filtering based on fuzzy mathematics methods is performed on the video image after watermark extraction to eliminate the effect of the watermark on the multimedia video data or video quality.

[0074] According to an embodiment of the present invention, using the method of dividing into blocks and groups for uniformly embedding the watermark in the multimedia video data, the amount of watermark information may be reduced, and the damage of the watermark to the quality of the multimedia video data may be greatly decreased. The accuracy for reflecting the multimedia video communication quality by the watermark is improved.

[0075] According to an embodiment of the present invention, the watermark is directly embedded in the original multimedia data before coding or in the video spatial domain, so that the watermark may completely equivalently reflect the intuitive quality of the multimedia. The accuracy of the multimedia video communication quality measurement is improved, and a quality loss degree caused by the multimedia coding process may be reflected.

[0076] According to an embodiment of the present invention, regarding the fitted PSNR as a quality measurement indication, which approximates a reference evaluation method, the accuracy of the objective quality evaluation for the multimedia video communication is improved.

[0077] According to an embodiment of the present invention, the effect of the watermark on the quality of the multimedia video data or video data may be reduced by performing smooth filtering post-processing for the watermark extraction, and the video communication quality may be ensured.

[0078] According to an embodiment of the present invention, there is provided a method for measuring multimedia video communication quality. The method includes: uniformly dividing multimedia video data frames into blocks of equal size, and embedding equal amount of the original watermark data in all of the blocks at the transmitting side, and transmitting the multimedia video data to a receiving side; extracting a recovered watermark from the received multimedia video data at the receiving side; determining a Pixel Recovery Rate of the recovered watermark relative to the original watermark in order to find a Peak Signal-to-Noise Ratio; and fitting to obtain an estimated value of the Peak Signal-to-Noise Ratio using a linear function of the Pixel Recovery Rate of the recovered watermark relative to the original watermark to measure the objective quality of the multimedia video communication, the pixel recovery rate referring to comparing distorted watermark with the original watermark to obtain a ratio of the number of correct watermark pixels after distortion to the total number of watermark pixels, wherein the linear function of the Pixel Recovery Rate of the recovered watermark relative to the original watermark is PSNR'=a₀ + a₁ × PRR, where a₀ and a₁ are determined according to a Mean Square Error criterion by using statistic data obtained during a previously performed multimedia video communication experiment.

[0079] According to the above-mentioned measures provided in the embodiments of the present invention, the quality of the multimedia video communication may be accurately reflected under the precondition that the video communication quality is not damaged. Rapid and real-time communication quality measurement may be provided in the video communication to facilitate positioning network faults and other system faults. It may facilitate the operators to display a service quality in real time and improve customer trust degree. It may facilitate the manufacturers to perform tests and validations during a developing process of a video communication product, so that the performance of the video communication product may be improved and video communication services may be popularized.

Brief Description of the Drawings

[0080] FIG.1 is a schematic diagram illustrating a principle of digital watermark technology;

[0081] FIG.2 is a block diagram illustrating a principle of a video communication quality measurement based on a digital watermark;

[0082] FIG.3 is a schematic diagram illustrating a image embedded with a watermark and watermark recovery situation;

[0083] FIG.4 is a flow chart of a method for measuring multimedia video communication quality according to an embodiment of the present invention;

[0084] FIG.5 is a schematic diagram illustrating uniformly dividing a frame into blocks of equal size according to an embodiment of the present invention;

[0085] FIG.6 (a) and (b) are schematic diagrams illustrating experimental results according to the embodiment 8 of the present invention.

Detailed Description of the Embodiments

[0086] Various objects, technical solution and advantages of one or some embodiments of the present invention may be more clearly understood with reference to the detailed description of the accompanying drawings and embodiments.

[0087] As described above, in view of the contradiction between a damaged video quality and an accurate reflection of the video communication quality, according to the embodiments of the present invention, there is provided a method for measuring video communication quality based on a digital watermark, which is able to balance the effect of a embedded watermark on video quality and the accuracy of reflecting the video quality, so that the video quality is not damaged and the video communication quality is accurately measured.

[0088] The principle of the method is: according to characteristics of compressed video, the watermark is completely and uniformly embedded in the video spatial domain using a watermark embedding algorithm based on dividing into blocks and groups, and the PSNR is fitted using the watermark recovery accuracy rate. The so-called fitting is a data processing method for computation mathematics: based on a series of discrete data points, a curve is obtained to represent these data points, wherein the total distance between the curve and these discrete data points is minimal according to a certain optimal criterion. Here, the fitting refers to approximating the PSNR curve with the PRR (Pixel Recovered Rate) data, the object of which is to show that both of them are very approximate and have very high correlativity and consistency, and accordingly provide quality evaluation indications applicable to compressed video with low code rate. Compared with the prior art, according to the embodiments of the present invention, the effect of the embedded digital watermark on the video quality is greatly reduced. The method of the embodiments of the present invention belongs to a Partial Reference Model method according to the above-mentioned quality evaluation categories. Without need of knowing the original video image, the method indirectly reflects the status of the original image using the original watermark image and the distorted watermark image which pass a transmission channel as a reference, and accordingly measures the video communication quality.

[0089] The objective quality measurement method for multimedia video communication based on a digital watermark according to the embodiments of the present invention includes three basic steps: embedding an original watermark in multimedia video data uniformly at a transmitting side, and transmitting the data to a receiving side; extracting a recovered watermark from the received multimedia video data through a process contrary to the process of watermark embedding at the receiving side; measuring objective quality of the multimedia video communication according to a distortion degree of the recovered watermark relative to the original watermark.

[0090] embodiment 1 of the present invention

[0091] FIG4 is a flow chart of an objective quality measurement method for multimedia video communication based on a digital watermark according to an embodiment of the present invention. Hereinafter the principle, implementation details and practical effect of each process will be described in detail.

[0092] With an example of video communication quality evaluation in the following description, the video data includes consecutive frame sequences. For the existing video, the B-DCT (Block-based DCT) coding and the motion compensated compression algorithm, such as H.263, H.264 and MPEG (Moving Picture Expert Group), are commonly used. Those skilled in the art may recognize that the method may be applicable to quality evaluation for other video communications with different coding or compression solutions and other multimedia video communications.

[0093] In Block 401, an original watermark image and a relevant algorithm or solution for video quality evaluation is determined. It needs to select a proper watermark at first for communication quality evaluation based on a digital watermark. In fact, selecting a proper digital watermark, which may not only reflect the video quality but also affect the video quality as little as possible, is important for the video communication quality evaluation.

[0094] embodiment 2 of the present invention

[0095] A bi-level image is selected in the embodiment, which is able to reduce the amount of information and appropriate for hiding watermark pixels. In addition, a black-and-white pattern, such as the watermark image shown in FIG.3, is also used. Black-and-white means 0-and-1, so that the amount of information carried by such an image is not large, but the image has high sensitivity and strong robustness for detecting error codes.

[0096] embodiment 3 of the present invention

[0097] A watermark image needs to be determined through a negotiation between both sides of communication, so in Embodiment 3 of the present invention, a transmitting side and a receiving side of the video communication negotiate through signaling communication to determine an original watermark and a specific solution for watermark embedment and watermark extraction.

[0098] After the original watermark is determined, the watermark will be uniformly embedded in a video image in accordance with the principle of the present invention. For the video compression coding standard of the B-DCT category, quantifying and coding are performed through a motion prediction, i.e. prediction and difference coding of a frame are performed with reference to anterior frames or posterior frames. A DCT transform is used for coding, so the transformed coding stream data is not completely equivalent to the former intuitive video image.

[0099] embodiment 4 of the present invention

[0100] In this embodiment, a digital watermark is embedded in the video spatial domain before a DCT transform, i.e. an original watermark is uniformly embedded in original multimedia data at a transmitting side, then compression coding and data packeting (for packet switching network) or framing (for circuit-switched network) are performed, and finally the data are transmitted to a receiving side via a network.

[0101] On one side, the effect of the video compression on the video quality may be reflected; and on the other side, large number of zero macro blocks exist in a compressed prediction frame or reverse prediction frame, and embedding the watermark in the transform domain will greatly reduce the coding efficiency and affect the coded video quality. Therefore, the method for embedding the watermark in the spatial domain has advantages on the above-mentioned two sides over the method for embedding the watermark in the transform domain.

[0102] preferred embodiment 5

[0103] In addition, according to a principle of visual psychics, compared with chrominance value, human eyes are not sensitive to a light change in a luminance value of an image. In order to reduce the loss of the video quality due to watermark embedment, in this embodiment, for a color image, an original watermark is embedded in a luminance component of the original multimedia data, i.e. a watermark signal is embedded in the luminance component (Y component) of a video image sequence.

[0104] In addition, for a multi-channel image (e.g. a multi-channel remote sensing image), an original watermark may be embedded in one or more channels of the original multi-channel image data.

[0105] In block 402 of Embodiment 1 of the present invention, each frame of the image is uniformly divided into blocks of equal size at the transmitting side, and equal amount of the original watermark data is embedded in a part of the blocks or all of the blocks.

[0106] Assuming that each frame includes M rows × N columns pixels in which the M rows are numbered from 0 to M-1 and the N columns are numbered from 0 to N-1. Each frame is uniformly divided into blocks according to the following solution:

[0107] each frame is divided into m strips numbered from 0 to m-1, the strip i(i = 0,1,L , m - 2) includes pixels from row i × k_v to row (i + 1)×k_v -1, and the strip m-1 includes pixels from row (m - 1)×k_v to row M-1, wherein m is a minimum integer which is greater than or equal to M/k_v ;

[0108] the strip i(i = 0,1,L , m-1) is divided into n blocks numbered from i×n to (i+1)×n-1, the block i×n+j (0,1,L , n - 2) includes pixels from column j×k_h to column (j+1)×k_h -1 of this strip, and the block n-1 includes pixels from column (n -1)×k_h to column N-1, wherein n is a minimum integer which is greater than or equal to N/k_h ;

[0109] equal amount of the original watermark data is embedded in each block of the obtained m × n blocks.

[0110] The specific details are shown in Fig. The size of the image in the video to be compressed is M × N, and the size of the watermark image is ┌M/k_v┐x┌N/k_h┐. Herein ┌ ┐ represents ceiling. One watermark pixel is embedded in each block whose size is k_v × k_h pixels in the original image, so that an error may be positioned and the effect of the watermark on the image quality is reduced. k_v represents a vertical direction period of the embedment, and k_h represents a horizontal direction period of the embedment.

[0111] In general, M or N is not divisible by k_v or k_h, the remainders of them are respectively r_v and r_h. Both of M and N are not divisible in a most complicated instance, thus four categories of areas exist in the top left part, on the right edge, on the lower edge and at the lower right corner of the image, which are marked with 1, 2, 3 and 4 in FIG5. The sizes of the blocks in the Area 1 are all k_v × k_h pixels. The sizes of the blocks in the Area 2 are all k_v × r_h pixels. The sizes of the blocks in the Area 3 are all r_v × k_h pixels. The size of the only one block in the Area 4 is r_v × r_h pixels.

[0112] The Area 1 is certain to exist, while the other areas may exist or not exist, depending on the situation of dividing. If the row number M is divisible by k_v, neither the Area 3 nor the Area 4 exists. Similarly, if the column N is divisible by k_h, neither the Area 2 nor the Area 4 exists.

[0113] By means of dividing into blocks, the watermark is distributed and embedded uniformly in each frame. A grouping embedding method is further needed to process different frames and embed the watermark in multiple frames. Especially for some instances, for example, when the communication bandwidth is very low or the selected watermark image is relatively large, it needs to further reduce the amount of the watermark data embedded in each frame of the image, i.e. it needs multiple frames to share a watermark.

[0114] In block 403, the video frame sequence is divided into groups at the transmitting side. Each group includes equal number of frames. At least two frames with equal interval in a same group are selected as the embedded frames, and equal number of blocks in each embedded frame are selected as the embedded blocks. Equal amount of the original watermark data are embedded in each embedded block.

[0115] To avoid significant effect on the quality and the code stream rate of a frame of an image due to embedding an integrate watermark image in each frame of the image to be compression-coded, the digital watermark image may be uniformly distributed in different frames. A part of the watermark is embedded in each frame of the image in the video sequence, or a sub-block is embedded in every other frame, etc., and the group of consecutive images is regarded as an embedding period. All the pixel points of the watermark are embedded in different positions in the image uniformly and dispersedly, and the watermark embedding positions of different images within a same embedding period are also different. In order to extract the embedded watermark conveniently at the decoding side, the position for embedding the watermark in each image within each embedding period is fixed.

[0116] Each frame is uniformly divided into L blocks numbered from 0 to L-1. The image may be uniformly divided into L strips horizontally or vertically, or uniformly divided in other ways. For example, the solution L = M × N shown in block 403. The method for embedding the original watermark also includes:

[0117] dividing the frame sequence into groups, each group including P × Q frames, from the frame 0 to the frame P × Q-1 of this group;

[0118] selecting the frame q × P (q = 0,1,L , Q-1) of each group as the embedded frame;

[0119] selecting the block q + r × Q (r = 0,L , R-1, and q + r × Q ≤ L) of the frame q × P (q = 0,1,L , Q-1) as the embedded block, wherein R is the maximum integer which is less than or equal to L/Q;

[0120] embedding equal amount of the original watermark data in all the embedded blocks of this group.

[0121] With an example of embedding with an interval, i.e. P = 2, the details are described hereinafter. Each group includes 2Q frames, R = └L/Q┘. The watermark is divided into Q blocks of equal size, and each block corresponds to one embedded frame. The frames 0, 2, 4, 6, 8,..., 2Q-2 of this group of image is divided into blocks, and the blocks are numbered with 0, 1, 2, 3,..., L-1 according to the order from the top to the bottom and from the left to the right.

[0122] In the frame 0 of this group, the blocks 0,Q,L , R×Q are regarded as the embedded blocks;

[0123] In the frame 2 of this group, the blocks 1,Q+1,L R×Q+1 are regarded as the embedded blocks; and the rest may be deduced analogically.

[0124] Till in the frame 2Q-2 of this group, the blocks Q-1,2Q-1,L , (R+1)×Q-1 are regarded as the embedded blocks.

[0125] It is noted that the above-mentioned block number may exceed L later, so the embedded blocks with those numbers need to be canceled. The frames 1,3,L ,2Q-1 in the period will not be processed.

[0126] Then in block 404, 1-bit watermark information is embedded in each embedded block. How to embed a watermark pixel in each block is a key factor affecting the performance of the algorithm of the embodiment of the present invention.

[0127] embodiment 6 of the present invention

[0128] In this embodiment, a method for embedding 1-bit data of the original watermark in one block of the multimedia video data includes:

[0129] setting an embedding strength as A = 2×Qp-1, wherein Q_P represents a quantization factor for the multimedia video communication compression coding ;

[0130] calculating an average luminance value µ of the block of the multimedia video data according to pixels;

[0131] modifying the luminance value of each pixel in the block of the data according to b' = b + c, wherein, b' is a modified value of the pixel, b is a value of the pixel before modification, and c is a modification amount;

[0132] the modification amount is calculated according to the following formula:

[0133] when the 1-bit original watermark data w = 1,

[0134] when the 1-bit original watermark data w = 0,

[0135] wherein γ is a remainder of µ being divided by 2A.

[0136] In block 405, a recovered watermark is extracted from the received video data at the receiving side through a process contrary to the process of watermark embedding. Corresponding to the watermark embedment, the watermark extraction is exactly a reverse process of the above-mentioned process: grouping and determining the embedded frames; determining the embedded blocks in each embedded frame; extracting the watermark information from each block. As for how to determine the embedded frames and further determine the embedded blocks, the details will not be described here. A method for extracting a watermark from an embedded block will be described.

[0137] embodiment 7 of the present invention

[0138] In this embodiment, a method for extracting 1-bit data of the recovered watermark from one block of the received video data based on the watermark embedding algorithm described in the embodiment 6 is:

[0139] setting an embedding strength as A = 2×Q_P-1, wherein Q_P represents the quantization factor for the multimedia video communication compression coding ;

[0140] calculating an average value µ' of the block of the received multimedia video data according to pixels;

[0141] calculating extracted data w' of the recovered watermark according to the following formula:

[0142] wherein γ' is a remainder of µ' being divided by 2A.

[0143] For all the embedded blocks in all the embedded frames in a group, a frame of integral watermark image may be recovered through recovering the watermark bit information.

[0144] In block 406, a PSNR is evaluated using a PRR (Pixel Recovered Rate) of the recovered watermark relative to the original watermark at the receiving side to measure the objective quality of the multimedia video communication.

[0145] The pixel recovery rate of the watermark is defined as PRR for evaluating the video quality by means of the watermark. The PRR refers to comparing the distorted watermark with the original watermark to obtain a ratio of the number of correct watermark pixels after distortion to the total number of the watermark pixels. I.e.,

[0146] PRR=100*(the number of the correct pixels in the distorted watermark/the total number of the watermark image pixels)%

[0147] Both video compression and error codes generated during a transmission process will affect the PRR value. Experiments show that although the PRR of the recovered watermark and the PSNR of the video quality to be evaluated are not completely identical, there is a fixed, even an approximately linear relationship between them. It is indicated that there is a very strong correlativity relationship between the quality evaluation result of the distorted video and the PSNR. Therefore, in a case that there is no reference video and the PSNR cannot be obtained, by means of establishing a relationship between the PRR and the PSNR, the PSNR of the distorted video is evaluated according to the PRR, and the object of evaluating the decoded and recovered video quality is achieved.

[0148] embodiment 8 of the present invention

[0149] Based on the embodiment 1, according to statistic data of a multimedia video communication experiment, the estimation value of the PSNR is fitted using a linear function of the PRR according to the MSE (Mean Square Error) criterion to measure the objective quality of the multimedia video communication in this embodiment.

[0150] Based on the embedding algorithm described above, the watermark information is embedded in a standard test sequence Claire (in QCIF format). Then, the video sequence embedded with the watermark is compressed according to a video compression standard H.263. The compressed video stream enters the decoder after channel "transmission" for video decoding. Finally, the embedded digital watermark is extracted from the decoded video and the PRR is calculated. A BSC (Binary Symmetric Channel) is used to simulate the channel for simplicity. The noise is uniformly distributed random noise. The average code error rate of the channel is 10^-5. The first diagram in FIG.6 shows the PSNR of the recovered video relative to the original video and the PRR of the watermark. It could be seen there is better consistency between the PSNR and the PRR.

[0151] In order to measure the similarity degree between the PSNR curve and the PRR curve, a normalized correlation efficiency of them in the diagram is calculated:

[0152] In the formula, X and Y respectively represent the PSNR and the PRR. Cov(X,Y) represents the covariance of X and Y. D(X) and D(Y) respectively represent the variances of X and Y.

[0153] Based on the experimental data, ρ =98.3%. It is obvious that there is a very strong correlativity between the PRR curve and the PSNR curve. The PSNR may be mapped by the PRR due to the strong correlativity between the PRR and the PSNR, i.e. a one-order function

[0154] is used to fit the actual PSNR. According to the MSE criterion, using the data in the diagram, when the mean square error

[0155] is the minimum value, a₀ = -16.9079, a₁ = 0.5632. The estimation PSNR' of the PSNR is calculated, as shown in FIG.6(a). Comparing the PSNR' with the PSNR, the MSE of them 0.735 is obtained. It could be seen that the PSNR' fits the PSNR very well.

[0156] In order to further validate the accuracy of the PSNR' as the PSNR estimation, the code error rate of the channel is changed to 1.5× 10^-5. The obtained PSNR of the recovered video and the estimation value PSNR' obtained through the PRR mapping are shown in FIG.6(b). The obtained correlation coefficient ρ of the two groups of data is 98.6%, and the MSE is only 0.696. It is obvious that the PSNR' fits the PSNR very well.

[0157] embodiment 9 of the present invention

[0158] In this embodiment, in order to further eliminate a loss of the video quality caused by the watermark embedment, the received video image will be post-processed after the recovered watermark is extracted to eliminate the effect of the watermark embedment. The post-processing may be smoothing filtering, median filtering, neural network filtering, mathematical morphology filtering or filtering based on fuzzy mathematics methods.

[0159] The watermark is embedded based on a block. If the embedding strength is too high, it will result in man-induced block effect and lead to decreased image quality. If the embedding strength is too low, it is difficult to resist the effect caused by video compression and results in a very low PRR of the recovered watermark. Thus in the embodiment of the present invention, a self-adaptive watermark embedding method is selected, and on restoring the watermark, smoothing filtering is performed for the image embedded with the watermark at the decoding side. An experiment shows that the effect of the watermark on the video objective quality may be effectively weakened in this way.

[0160] embodiment 10 of the present invention

[0161] In this embodiment, the performance of the method for measuring the multimedia video communication quality based on a digital watermark according to the embodiment of the present invention is evaluated through experiments. The results are shown in Table 1 and Table 2.

[0162] Table 1 and Table 2 respectively show the effect on the coded video quality and the bit rate caused by watermark embedding using the method of the embodiment of the present invention. It could be seen that the effect on the multimedia video data quality brought by the method of the embodiment of the present invention is little.

[0163] Table 1 at the same bit rate, effect on coded video quality caused by the watermark embedment (the frame rate is 30 frames/s)

Bit rate (bps)	20K	50K	80K
PSNR before embedding (dB)	37.60	41.51	43.87
PSNR after embedding (dB)	37.44	41.42	43.80

[0164] Table 2with the same quantization factor, effect on original video quality and bit rate caused by the watermark embedment

Quantization step length		QP=5	QP=8	QP=12
Before embedding	Bit rate (bps)	66.67K	33.60K	20.23K
	PSNR (dB)	40.44	37.59	35.31
After embedding	Bit rate (bps)	69.60K	35.97K	21.02K
	PSNR (dB)	40.37	37.49	35.24

[0165] Those skilled in the art may recognize that the technical solution of the embodiment of the present invention also may be used in conjunction with quality measurement methods with reference and achieve better effect through information fusion, etc.

[0166] While the present invention has been illustrated and described with reference to some preferred embodiments, those skilled in the art may recognize that various variations and modifications may be made without departing from the scope of the present invention as defined by the accompanying claims.

Claims

1. A method for measuring multimedia video communication quality, wherein the method comprises:

uniformly dividing multimedia video data frames into blocks of equal size, and embedding equal amount of the original watermark data in all of the blocks at the transmitting side, and transmitting the multimedia video data to a receiving side;

extracting a recovered watermark from the received multimedia video data at the receiving side;

determining a Pixel Recovery Rate of the recovered watermark relative to the original watermark in order to find a Peak Signal-to-Noise Ratio; characterized by

obtaining an estimated value of the Peak Signal-to-Noise Ratio using a linear function of the Pixel Recovery Rate of the recovered watermark relative to the original watermark to measure the objective quality of the multimedia video communication, the pixel recovery rate referring to a ratio of the number of correct watermark pixels after distortion to the total number of watermark pixels, said ratio being obtained by comparing the distorted watermark with the original watermark, wherein

said linear function of said Pixel Recovery Rate is


where a₀ and a₁ are determined according to a Mean Square Error criterion by using statistic data obtained during a multimedia video communication experiment, PSNR' referring to said estimated Peak Signal-to-Noise Ratio and PRR referring to said Pixel Recovery Rate.

2. The method of claim 1, further comprising:

before transmitting the data to the receiving side, performing compression coding or data packaging or framing at the transmitting side.

3. The method of claim 1, further comprising: dividing the multimedia video data into groups at the transmitting side, each group comprising equal number of frames, and selecting at least two frames with equal interval in a same group as embedded frames, uniformly dividing each of the embedded frame into a plurality of blocks of equal size, selecting equal number of blocks from the plurality of blocks as embedded blocks, and embedding equal amount of the original watermark data in each of the embedded block.

4. The method of claim 1, wherein, each multimedia video data frame comprises M rows × N columns pixels in which the M rows are numbered from 0 to M-1 and the N columns are numbered from 0 to N-1, and uniformly dividing a multimedia video data frame into blocks of equal size comprises:

dividing each multimedia video data frame into m strips numbered from 0 to m-1, the strip i (i = 0,1,...,m - 2) comprising pixels from row i × k_v to row (i+1)×k_v-1, and the strip m-1 comprising pixels from row (m-1)×k_v to row M-1, wherein m is a minimum integer greater than or equal to M/k_v ; if m = M/k_v, the strip m-1 comprising zero row of pixels;

dividing the strip i (i = 0,1, ... , m-1) into n blocks numbered from i×n to (i+1)×n-1, the block i×n+j (0,1,..., n-2) comprising pixels from column j×k_h to column (j+1)×k_h-1 of this strip, and the block n-1 comprising pixels from column (n-1)×k_h to column N-1, wherein n is a minimum integer greater than or equal to N/k_h ; if n = N/k_h, the strip n-1 comprising zero column of pixels;

embedding equal amount of the original watermark data in each of the obtained m×n blocks.

5. The method of claim 3, wherein, each frame of the multimedia video data is uniformly divided into L blocks numbered from 0 to L-1, and the embedding the original watermark further comprises:

dividing a frame sequence into groups, each group comprising P × Q frames, which sequentially are frame 0 to frame P×Q-1 of the group;

selecting a frame q×P(q = 0,..., Q-1) of each group as the embedded frame;

selecting a block q+r×Q(r = 0, ... , R-1, and q + r × Q ≤ L) of the embedded frame q × P (q = 0, ... , Q-1) as the embedded block, wherein R is a maximum integer less than or equal to L/Q ;

embedding equal amount of the original watermark data in all the embedded blocks of the group.

6. The method of claim 1, further comprising: determining a bit number of the original watermark data embedded in all of the blocks according to the size of the data block.

7. The method of claim 6, wherein, embedding 1-bit data of the original watermark in the multimedia video data block comprises:

calculating an average value µ of the multimedia video data block according to pixels;

modifying a value of each pixel in the multimedia video data block according to b' = b + c , wherein, b' is a modified value of a pixel, b is a value of the pixel before modification, and c is a modification amount;
wherein the modification amount is calculated according to the following formula:

when the original watermark data bit is 1,

when the original watermark data bit is 0,

wherein γ is a remainder of µ being divided by 2A, A represents an embedding strength and is equal to 2×Q_P-1, and Q_P represents a quantization factor for the multimedia video communication compression coding.

8. The method of claim 6, wherein, extracting a data bit of the recovered watermark from the received multimedia video data block comprises:

calculating an average value µ' of the received multimedia video data block according to pixels;

calculating extracted data w' of the recovered watermark according to the following formula:


wherein γ' is a remainder of µ' being divided by 2A, A represents an embedding strength and is equal to 2×Q_P-1, and Q_P represents a quantization factor for the multimedia video communication compression coding.

9. The method of claim 1, wherein a₀ is determined as -16.9079, and a₁ is determined as 0.5632.

10. The method of claim 1, further comprising:

on extracting the recovered watermark, post-processing the received multimedia video data to eliminate effect of the watermark embedment, wherein the post-processing comprises one of the following:

smoothing filtering, median filtering, neural network filtering, mathematical morphology filtering and filtering based on fuzzy mathematics methods.

11. The method of claim 1, further comprising:

determining the original watermark, relevant parameters and a strategy for embedding the original watermark and extracting the recovered watermark through a communication negotiation between the transmitting side and the receiving side.

12. The method of claim 1, further comprising: determining the original watermark, the relevant parameters and the strategy for embedding the original watermark and extracting the recovered watermark at the transmitting side, and informing the receiving side before communication.

13. The method of claim 2, wherein, for a color image, the original watermark is embedded in a luminance component of the original multimedia data;
for a multi-channel image, the original watermark is embedded in one or more channels of the original multi-channel image data.

14. The method of any of claims 1, 3, 5, 8-13, wherein the original watermark is a bi-level image.

Ansprüche

1. Ein Verfahren zum Messen von Multimedia-Video-Kommunikationsqualität, wobei das Verfahren aufweist:

gleichmäßiges Aufteilen von Multimedia-Videodatenrahmen in Blöcke gleicher Größe und Einbetten der gleichen Menge der ursprünglichen Wasserzeichendaten in alle Blöcke auf der Sendeseite, und Senden der Multimedia-Videodaten an eine Empfangsseite,

Extrahieren eines wiederhergestellten Wasserzeichens aus den empfangenen Multimedia-Videodaten auf der Empfängerseite,

Bestimmen einer Bildpunkt-Wiederherstellungsrate des wiederhergestellten Wasserzeichens in Bezug auf das ursprüngliche Wasserzeichen zum Ermitteln eines Spitzen-Signal-Rausch-Verhältnisses, gekennzeichnet durch

das Erhalten eines geschätzten Wertes des Spitzen-Signal-Rausch-Verhältnisses mittels einer linearen Funktion der Bildpunkt-Wiederherstellungsrate des wiederhergestellten Wasserzeichens in Bezug auf das ursprüngliche Wasserzeichen zum Messen der objektiven Qualität der Multimedia-Video-Kommunikation, wobei sich die Bildpunkt-Wiederherstellungsrate auf ein Verhältnis der Anzahl von korrekten Wasserzeichen-Bildpunkten nach der Verzerrung zu der gesamten Anzahl von Wasserzeichen-Bildpunkten bezieht, wobei das Verhältnis durch Vergleichen des verzerrten Wasserzeichens mit dem ursprünglichen Wasserzeichen erhalten wird, wobei die lineare Funktion der Bildpunkt-Wiederherstellungsrate


wobei a₀ und a₁ gemäß einem Mittlere-Quadratische-Abweichung-Kriterium mittels während eines Multimedia-Video-Kommunikationsversuchs erhaltener statistischer Daten bestimmt werden, wobei PSNR' sich auf das geschätzte Spitzen-Signal-Rausch-Verhältnis bezieht und PRR sich auf die Bildpunkt-Wiederherstellungsrate bezieht.

2. Das Verfahren gemäß Anspruch 1, ferner aufweisend:

Durchführen eines Kompressionskodierens oder eines Datenpackens oder eines Rahmens auf der Sendeseite vor dem Senden der Daten an die Empfangsseite.

3. Das Verfahren gemäß Anspruch 1, ferner aufweisend: Aufteilen der Multimedia-Videodaten in Gruppen auf der Sendeseite, wobei jede Gruppe eine gleiche Anzahl von Rahmen aufweist, und Auswählen von mindestens zwei Rahmen mit gleichem Intervall in einer gleichen Gruppe als eingebettete Rahmen, gleichmäßiges Aufteilen jedes eingebetteten Rahmens in eine Mehrzahl von Blöcken gleicher Größe, Auswählen einer gleichen Anzahl von Blöcken aus der Mehrzahl von Blöcken als eingebettete Blöcke, und Einbetten einer gleichen Menge der ursprünglichen Wasserzeichendaten in jeden eingebetteten Block.

4. Das Verfahren gemäß Anspruch 1, wobei jeder Multimedia-Videodatenrahmen M Zeilen x N Spalten von Bildpunkten aufweist, wobei die M Zeilen von 0 bis M-1 nummeriert sind und die N Spalten von 0 bis N-1 nummeriert sind, und wobei das gleichmäßige Aufteilen eines Multimedia-Videodatenrahmens in Blöcke gleicher Größe aufweist:

Aufteilen jedes Multimedia-Videodatenrahmens in m von 0 bis m-1 nummerierte Streifen, wobei der Streifen i (i = 0, 1, ... , m - 2) Bildpunkte von Reihe ixk_v bis Reihe (i + 1) x k_v - 1 aufweist, und wobei der Streifen m - 1 Bildpunkte von Reihe (m - 1) x k_v bis Reihe M-1 aufweist, wobei m eine minimale ganze Zahl größer oder gleich M/k_v ist, wobei der Streifen m -1 null Pixelreihen aufweist, wenn m = M/k_V,

Aufteilen des Streifens i(i = 0, 1, ... , m-1) in n von ixn bis (i + 1) x n - 1 nummerierte Blöcke, wobei der Block ixn + j (0, 1, ... , n-2) Bildpunkte von Spalte j x k_h bis Spalte (j+ 1) x k_h -1 des Streifens aufweist, und wobei der Block n - 1 Bildpunkte von Spalte (n - 1) x k_h bis Spalte N-1 aufweist, wobei n eine minimale ganze Zahl größer oder gleich N/k_h ist, wobei der Streifen n-1 null Spalten von Bildpunkten aufweist, wenn n= N/k,

Einbetten der gleichen Menge ursprünglicher Wasserzeichendaten in jeden der erhaltenen mxn Blöcke.

5. Das Verfahren gemäß Anspruch 3, wobei jeder Rahmen der Multimedia-Videodaten gleichmäßig in L von 0 bis L-1 nummerierte Blöcke aufgeteilt wird, und das Einbetten des ursprünglichen Wasserzeichens ferner aufweist:

Aufteilen einer Rahmensequenz in Gruppen, wobei jede Gruppe P x Q Rahmen aufweist, die nacheinander Rahmen 0 bis Rahmen P x Q - 1 der Gruppe sind,

Auswählen eines Rahmens q x P (q = 0, ... , Q-1) jeder Gruppe als den eingebetteten Rahmen,

Auswählen eines Blocks q + r x Q (r = 0, ... , R-1, und q + r x Q ≤ L) des eingebetteten Rahmens q x P (q = 0, ... , Q-1) als den eingebetteten Block, wobei R eine maximale ganze Zahl kleiner oder gleich L/Q ist,

Einbetten der gleichen Menge ursprünglicher Wasserzeichendaten in alle eingebetteten Blöcke der Gruppe.

6. Das Verfahren gemäß Anspruch 1, ferner aufweisend: Bestimmen einer Bitanzahl der in alle Blöcke eingebetteten ursprünglichen Wasserzeichendaten gemäß der Größe des Datenblocks.

7. Das Verfahren gemäß Anspruch 6, wobei das Einbetten von 1-Bit-Daten des ursprünglichen Wasserzeichens in den Multimedia-Videodatenblock aufweist:

Berechnen eines Durchschnittswertes µ des Multimedia-Videodatenblocks gemäß Bildpunkten,

Modifizieren eines Wertes jedes Bildpunktes in dem Multimedia-Videodatenblock gemäß b' = b + c, wobei b' ein modifizierter Wert eines Bildpunktes ist, b ein Wert des Bildpunktes vor der Modifizierung ist, und c eine Modifizierungsmenge ist,
wobei die Modifizierungsmenge gemäß der folgenden Formel berechnet wird:

wenn das ursprüngliche Wasserzeichen-Datenbit 1 ist:

wenn das ursprüngliche Wasserzeichen-Datenbit 0 ist:

wobei γ ein Rest von µ geteilt durch 2A ist, A eine Einbettungsstärke repräsentiert und gleich 2 x Q_P - 1 ist, und Q_P einen Quantisierungsfaktor für das Multimedia-Video-Kommunikations-Kompressionskodieren repräsentiert.

8. Das Verfahren gemäß Anspruch 6, wobei das Extrahieren eines Datenbits des wiederhergestellten Wasserzeichens aus dem empfangenen Multimedia-Videodatenblock aufweist:

Berechnen eines Durchschnittswertes µ' des empfangenen Multimedia-Videodatenblocks entsprechend Bildpunkten,

Berechnen von extrahierten Daten w' des wiederhergestellten Wasserzeichens gemäß der folgenden Formel:


wobei γ' ein Rest von µ' geteilt durch 2A ist, A eine Einbettungsstärke repräsentiert und gleich 2 x Qp - 1 ist, und Q_P einen Quantisierungsfaktor für das Multimedia-Videokommunikations-Kompressionskodieren repräsentiert.

9. Das Verfahren gemäß Anspruch 1, wobei a₀ als -16,9079 bestimmt wird und a₁ als 0,5632 bestimmt wird.

10. Das Verfahren gemäß Anspruch 1, ferner aufweisend:

beim Extrahieren des wiederhergestellten Wasserzeichens das Nachbearbeiten der empfangenen Multimedia-Videodaten zum Beseitigen der Wirkung des Wasserzeichen-Einbettens, wobei das Nachbearbeiten eines aus den folgenden aufweist:

Glättungsfiltern, Medianfiltern, Neuronen-Netzwerkfiltern, Mathematische-Morphologie-Filtern und Filtern auf der Grundlage von Unscharfe-Mathematik-Methoden.

11. Das Verfahren gemäß Anspruch 1, ferner aufweisend:

Bestimmen des ursprünglichen Wasserzeichens, relevanter Parameter und einer Strategie zum Einbetten des ursprünglichen Wasserzeichens und Extrahieren des wiederhergestellten Wasserzeichens durch ein Kommunikationsverhandeln zwischen der Sendeseite und der Empfangsseite.

12. Das Verfahren gemäß Anspruch 1, ferner aufweisend: Bestimmen des ursprünglichen Wasserzeichens, der relevanten Parameter und der Strategie zum Einbetten des ursprünglichen Wasserzeichens und Extrahieren des wiederhergestellten Wasserzeichens auf der Sendeseite, sowie Benachrichtigen der Empfängerseite vor der Kommunikation.

13. Das Verfahren gemäß Anspruch 2, wobei für ein Farbbild das ursprüngliche Wasserzeichen in eine Luminanzkomponente der ursprünglichen Multimediadaten eingebettet wird, und
für ein Mehrkanalbild das ursprüngliche Wasserzeichen in einen oder mehrere Kanäle der ursprünglichen Mehrkanalbilddaten eingebettet wird.

14. Das Verfahren gemäß einem der Ansprüche 1, 3, 5, 8 bis 13, wobei das ursprüngliche Wasserzeichen ein Zwei-Ebenen-Bild ist.

Revendications

1. Procédé de mesure d'une qualité d'une communication vidéo multimédia,
dans lequel le procédé comporte les étapes suivantes :

diviser de manière uniforme des trames de données vidéo multimédia en blocs de taille égale, insérer dans tous les blocs au niveau du côté émetteur une quantité égale de données d'un filigrane numérique original, et transmettre les données vidéo multimédia à un côté récepteur ;

extraire un filigrane récupéré des données vidéo multimédia reçues au niveau du côté récepteur ;

déterminer une vitesse de récupération de pixels du filigrane récupéré par rapport au filigrane original afin de trouver un rapport signal/bruit de crête ; caractérisé par

l'obtention d'une valeur estimée du rapport signal/bruit de crête à l'aide d'une fonction linéaire d'un taux de récupération de pixels du filigrane récupéré par rapport au filigrane original pour mesurer la qualité objective de la communication vidéo multimédia, le taux de récupération de pixels correspond à un rapport du nombre de pixels de filigrane corrects après distorsion au nombre total de pixels du filigrane, ledit rapport étant obtenu en comparant le filigrane distordu avec le filigrane original, dans lequel

ladite fonction linéaire de ladite vitesse de récupération de pixels est


où a₀ et a₁ sont déterminés en fonction d'un critère d'erreur quadratique moyenne en utilisant les données statistiques obtenues pendant une expérience de communication vidéo multimédia, PSNR' désignant ledit rapport signal/bruit de crête estimé et PRR désignant ledit taux de récupération de pixels.

2. Procédé selon la revendication 1, comprenant en outre l'étape, consistant à :

à effectuer un codage par compression, ou une mise en paquet ou un verrouillage de trames de données au niveau du côté émetteur, avant de transmettre les données au côté récepteur.

3. Procédé selon la revendication 1, comprenant en outre :

une division des données vidéo multimédia en groupes au niveau du côté émetteur, chaque groupe comprenant un nombre égal de trames, et une sélection d'au moins deux trames avec un intervalle égal dans un même groupe

en tant que trames insérées, une division uniforme de chacune des trames insérées en une pluralité de blocs de taille égale, une sélection d'un nombre égal de blocs de la pluralité de blocs en tant que trames insérées, et une insertion d'une quantité égale des données du filigrane original dans chacun des blocs insérés.

4. Procédé selon la revendication 1, dans lequel chaque trame de données vidéo multimédia comprend des pixels sur M rangées x N colonnes dans lequel les M rangées sont numérotées de 0 à M - 1 et les N colonnes sont numérotées de 0 à N - 1, et une division uniforme d'une trame de données vidéo multimédia en blocs de taille égale comprenant :

division de chaque trame de données vidéo multimédia en m bandes numérotées de 0 à m - 1, la bande i (i = 0,1,..., m - 2) comprenant des pixels de la rangée i x k_v à la rangée (i + 1) x k_v - 1, et la bande m - 1 comprenant des pixels de la rangée (m - 1) x k_v à la rangée M - 1, dans laquelle m est un entier minimum supérieur ou égal à M/k_v ; si m = M/k_v, la bande m - 1 comprend zéro rangée de pixels ;

division de la bande i(i = 0,1,..., m - 1) en n blocs numérotés de i x n à (i + 1) x n - 1, le bloc i x n + j (0,1,..., n - 2) comprenant des pixels de la colonne j x k_h à la colonne (j + 1) x k_h - 1 de cette bande, et le bloc n - 1 comprenant des pixels de la colonne (n - 1) x k_h à la colonne N - 1, dans lequel n est un entier minimum supérieur ou égal à N/k_h; si n = N/k_h, la bande n - 1 comprend zéro colonne de pixels ;

insertion d'une quantité égale de données du filigrane original dans chacun des blocs m x n obtenus.

5. Procédé selon la revendication 3, dans lequel chaque trame des données vidéo multimédia est divisée de manière uniforme en L blocs numérotés de 0 à L - 1, et l'insertion du filigrane original comprenant en outre :

division d'une séquence de trames en groupes, chaque groupe comprenant P x Q trames qui, de manière séquentielle, vont de la trame 0 à la trame P x Q - 1 du groupe ;

sélection d'une trame q x P (q = 0,..., Q - 1) de chaque groupe en tant que trame insérée ;

sélection d'un bloc q + r x Q (r = 0, ... R - 1, et q + r x Q ≤ L) de la trame insérée q x P (q = 0,..., Q - 1) en tant que trame insérée, dans lequel R est un entier maximum inférieur ou égal à L/Q ;

insertion d'une quantité égale des données du filigrane original dans tous les blocs insérés du groupe.

6. Procédé selon la revendication 1, comprenant en outre : la détermination d'un nombre binaire de données du filigrane original insérées dans tous les blocs en fonction de la taille des blocs de données.

7. Procédé selon la revendication 6, dans lequel l'insertion de données 1 bit du filigrane original dans les données vidéo multimédia comporte :

le calcul d'une valeur moyenne µ du bloc de données vidéo multimédia en fonction des pixels ;

la modification d'une valeur de chaque pixel dans le bloc de données vidéo multimédia en fonction de b' = b + c, où b' est une valeur modifiée d'un pixel, b est une valeur du pixel avant modification, et c est une quantité de modification ;
dans laquelle la quantité de modification est calculée en fonction de la formule suivante :

lorsque le bit de données du filigrane original est égal à 1,

lorsque le bit de données du filigrane original est égal à 0,

où γ est un reste de µ divisé par 2A, A représente une force d'insertion et est égal à 2 x Q_p - 1, et Q_p représente un facteur de quantification pour le codage par compression de la communication vidéo multimédia.

8. Procédé selon la revendication 6, dans lequel l'extraction d'un bit de données du filigrane récupéré depuis le bloc de données vidéo multimédia reçues comporte :

le calcul d'une valeur moyenne µ' du bloc de données vidéo multimédia reçues en fonction des pixels ;

le calcul des données extraites ω' du filigrane récupéré en fonction de la formule suivante :


dans laquelle γ' est un reste de µ' divisé par 2A, A représente une force d'insertion et est égal à 2 x Q_p - 1, et Q_p représente un facteur de quantification pour le codage par compression de la communication vidéo multimédia.

9. Procédé selon la revendication 1, dans lequel a₀ est défini comme étant égal à -16,9079, et a₁ est défini comme étant égal à 0,5632.

10. Procédé selon la revendication 1, comprenant en outre :

le traitement ultérieur des données vidéo multimédia reçues pour éliminer l'effet de l'insertion du filigrane, lors de l'extraction du filigrane récupéré, dans lequel le post-traitement comporte une des opérations suivantes :

filtrage de lissage, filtrage médian, filtrage de réseau neural, filtrage de morphologie mathématique et filtrage reposant sur des procédés mathématiques flous.

11. Procédé selon la revendication 1, comprenant en outre :

la détermination du filigrane original, les paramètres pertinents et une stratégie d'insertion du filigrane original et l'extraction du filigrane récupéré à travers une négociation de communication entre le côté émetteur et le côté récepteur.

12. Procédé selon la revendication 1, comprenant en outre : la détermination du filigrane original, les paramètres pertinents et la stratégie d'insertion du filigrane original et l'extraction du filigrane récupéré au niveau du côté émetteur, et informer le côté récepteur avant la communication.

13. Procédé selon la revendication 2, dans lequel, pour une image en couleur, le filigrane original est inséré dans une composante de luminance des données multimédia originales ;
pour une image multi-canaux, le filigrane original est inséré dans un ou plusieurs canaux des données d'image multi-canaux originales.

14. Procédé selon l'une des revendications 1, 3, 5, 8 à 13, dans lequel le filigrane original est une image à deux niveaux.

Drawing