(19)
(11)EP 2 553 626 B1

(12)EUROPEAN PATENT SPECIFICATION

(45)Mention of the grant of the patent:
29.07.2020 Bulletin 2020/31

(21)Application number: 11768528.9

(22)Date of filing:  25.03.2011
(51)International Patent Classification (IPC): 
G06K 7/10(2006.01)
G06K 9/46(2006.01)
G06K 9/18(2006.01)
(86)International application number:
PCT/IB2011/001156
(87)International publication number:
WO 2011/128777 (20.10.2011 Gazette  2011/42)

(54)

SEGMENTATION OF TEXTUAL LINES IN AN IMAGE THAT INCLUDE WESTERN CHARACTERS AND HIEROGLYPHIC CHARACTERS

SEGMENTIERUNG VON TEXTZEILEN IN EINEM BILD MIT WESTLICHEN ZEICHEN UND HIEROGLYPHEN

SEGMENTATION DE LIGNES TEXTUELLES DANS UNE IMAGE COMPRENANT DES CARACTÈRES OCCIDENTAUX ET DES CARACTÈRES HIÉROGLYPHIQUES


(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: 31.03.2010 US 751309

(43)Date of publication of application:
06.02.2013 Bulletin 2013/06

(73)Proprietor: Microsoft Technology Licensing, LLC
Redmond, WA 98052 (US)

(72)Inventor:
  • MITIC, Ivan
    Redmond, WA 98052-6399 (US)

(74)Representative: Goddar, Heinz J. 
Boehmert & Boehmert Anwaltspartnerschaft mbB Pettenkoferstrasse 22
80336 München
80336 München (DE)


(56)References cited: : 
US-A1- 2004 146 216
US-A1- 2008 310 721
US-A1- 2010 067 794
US-A1- 2005 027 511
US-A1- 2009 136 135
US-A1- 2010 074 525
  
  • LU Y ED - IMPEDOVO SEBASTIANO LIU CHENG-LIN IMPEDOVO DONATO PIRLO GIUSEPPE: "Machine printed character segmentation--an overview", PATTERN RECOGNITION, ELSEVIER, GB, vol. 28, no. 1, 1 January 1995 (1995-01-01), pages 67-80, XP004014034, ISSN: 0031-3203, DOI: 10.1016/0031-3203(94)00068-W
  • SENI G ET AL: "SEGMENTING HANDWRITTEN TEXT LINES INTO WORDS USING DISTANCE ALGORITHMS", MACHINE VISION APPLICATIONS IN CHARACTER RECOGNITION AND INDUSTRIAL INSPECTION. SAN JOSE, FEB. 10 - 12, 1992; [MACHINE VISION APPLICATIONS IN CHARACTER RECOGNITION AND INDUSTRIAL INSPECTION], BELLINGHAM, SPIE, US, 1 January 1992 (1992-01-01), pages 61-72, XP000633702, ISBN: 978-0-8194-0815-0
  
Note: Within nine months from the publication of the mention of the grant of the European patent, any person may give notice to the European Patent Office of opposition to the European patent granted. Notice of opposition shall be filed in a written reasoned statement. It shall not be deemed to have been filed until the opposition fee has been paid. (Art. 99(1) European Patent Convention).


Description

Background of the Invention



[0001] Optical character recognition (OCR) is a computer-based translation of an image of text into digital form as machine-editable text, generally in a standard encoding scheme. This process eliminates the need to manually type the document into the computer system. An OCR process typically begins by obtaining an electronic file of a physical document bearing the printed text message and scanning the document with a device such as an optical scanner. Such devices produce an electronic image of the original document. The output image is then supplied to a computer or other processing device and processes the image of the scanned document to differentiate between images and text and determine what letters are represented in the light and dark areas, as is e.g. disclosed in US2004/0146216.

[0002] As a result of the increasing use of computers and the Internet, coupled with the more frequent usage of English language around the world, it has become common to find textual images that include a combination of Western words and East Asian (e.g., Chinese, Japanese, Korean) text, often in the form of Western Words mixed within a selection of East Asian text. Accordingly, an OCR engine that is to be used with East Asian text should ideally be able to recognize a textual line with a mix of East Asian and Western text.

Summary



[0003] In order to support the OCR process, East Asian or hieroglyphic text textual lines with only East Asian text inter-character breaks and individual characters need to be recognized. In addition, for textual lines with both Western and East Asian text the Western and East Asian fragments of text need to be separated from one another and the appropriate text recognition logic needs to be applied to each fragment.

[0004] The invention is defined in the independent claims. Specific embodiments are defined in the dependent claims.

Brief Description of the Drawings



[0005] 

FIGs. 1 and 2 show a textual line of Western and East Asian text, respectively.

FIG. 3 shows a text line that contains a mix of Western and hieroglyphic text.

FIG. 4 shows one example of an image processing apparatus that performs the process of segmenting Western and hieroglyphic portions of textual lines.

FIGs. 5a and 5b show examples of characters of East Asian text surrounded by their respective bounding boxes and the candidate inter-character breaks between them.

FIG. 6a shows an image of a character prior to pre-processing along with the width and height of its bounding box; FIG. 6b shows the character after it has been stretched.

FIGs. 7-12 show a textual line as it undergoes the text segmentation process.

FIG. 13 is a flowchart illustrating one example of a process of dividing textual lines into Western and hieroglyphic text segments.


Detailed Description



[0006] One important aspect of the optical character recognition (OCR) process is line recognition and segmentation. However, the concept of a line has a different meaning for Western text and East Asian text (or more generally, any hieroglyphic-based in which a single written character represents a word). This distinction can be seen in FIGs. 1 and 2, which show a textual line of Western and East Asian text, respectively. In particular, in the case of a hieroglyphic text line there is only the concept of characters while in a case of a Western text line there is also the concept of words. In other cases, such as those being addressed herein, a text line contains a mix of Western and hieroglyphic text (FIG. 3).

[0007] In order to support the OCR process for a mix of Western and East Asian or hieroglyphic text a number of problems need to be addressed. In particular, for textual lines with only East Asian text correct inter-character breaks and individual characters need to be recognized. In addition, for textual lines with both Western and East Asian text the textual line must be properly segmented into Western and East Asian segments of text. Each of these problems will be addressed in turn.

[0008] FIG. 4 shows one example of an image processing apparatus 100 that may perform the process of segmenting Western and hieroglyphic portions of textual lines. The apparatus, which may be incorporated in an OCR engine, can be used by the OCR engine to identify Western and East Asian or other hieroglyphic characters in a textual line. The apparatus includes an input component 102 for receiving an input image that includes at least one textual line. An inter-character break identifier component 104 identifying candidate inter-character breaks along a textual line and an inter-character break classifier component 106 classifies each of the candidate inter-character breaks as an actual break, a non-break or an indeterminate break. The apparatus also includes a character recognition engine 108 for recognizing the candidate characters based at least in part on a feature set extracted from each respective candidate character. The character recognition engine 108 includes a character pre-processing component 110, a character feature extraction component 112 and a character classification component 113. A western and hieroglyphic text classifier component 114 segments textual lines into Western text segments and an East Asian or other hieroglyphic text segments. The apparatus 100 also includes an output component 116 that receives the results from the Western and hieroglyphic text classifier component and generates the information concerning the text-lines in a form that allows it to be employed by subsequent components of the OCR engine.

[0009] Each of the aforementioned components will be discussed below.

Inter-character Breaks



[0010] The inter-character break identifier component 104 can identify candidate breaks between East Asian or hieroglyphic characters using well-known vertical projection techniques, which may include techniques for separating symbols that touch one another.. FIGs. 5a and 5b show examples of inter-character breaks in East Asian text. Not all candidate inter-character breaks identified in this manner will be true breaks, however. For instance, there may be no clear spatial distinction between individual characters. Conversely, there may appear to be a small spatial distinction within an individual character that may be misidentified as an inter-character break. Accordingly, a method is needed to determine which candidate inter-character breaks are most likely actual inter-character breaks.

[0011] Each candidate inter-character break is classified by the inter-character break classifier component 106 as a break point, a non-break point or an indeterminate or undecided point. A break point is an actual break between two characters and a non-break point is located within an individual character. Accordingly, a non-break point is not a true inter-character break. Stated differently, a character cannot span across a break point and a character cannot start or end at a non-break point. However, a character can span across, start at, or end at an indeterminate point.

[0012] In order to classify a candidate inter-character break, each candidate's probability is computed and two thresholds are set, one for a break, and one for a non-break. If p(x) is the probability that a candidate x represents an actual break, than if p(x) > BREAK, x will be classified as a break, if p(x) < NON_BREAK, x will be classified as a non-break, and when NON_BREAK <= p(x) <= BREAK the candidate is classified as an undecided point, which can be classified in later stages of the processing.

[0013] There are a number of advantages that arise from the use of the aforementioned inter-character break classification scheme. For instance, all inter-character candidate breaks classified as non-breaks are removed in subsequent processing steps, resulting in better performance and accuracy. In addition, all inter-character candidate breaks classified as breaks can only be treated as a point at which a character starts or ends, which also results in better performance and accuracy.

[0014] The following terminology will be useful in establishing characteristics or features of inter-character break candidates that may be used to classify them as a break point, a non-break point or an undecided point. These features, including character bounding boxes, may be determined by the inter-character break classifier component 106.
Bi - ith break. Each break is defined with two x coordinates (Bi.start and Bi.end) and
Bi.size = Bi.end - Bi.start + 1.
BBpi - bounding box preceding i-th inter-character break (the rectangular bounding boxes are visible in FIGs. 5a and 5b)
BBsi - bounding box succeeding i-th inter-character break
BBsi.top and BBsi.bottom - top and bottom coordinates, respectively, of bounding box BBsi
BBpi.top and BBpi.bottom - top and bottom coordinates, respectively, of bounding box BBpi
BBpi.width, BBpi.height - width and height, respectively of bounding box BBpi
BBsi.width, BBps.height - width and height, respectively of bounding box BBsi
MBS - median break size for the given line
ECH - estimated character height for the given line (or median bounding box height if there is no better estimate)
ECW - estimated character width (which can be estimated along with ECH if there is no better estimate)
Bpi - index of preceding break which x coordinate is closest to BBpi.right - ECW
Bsi - index of succeeding break which x coordinate is closest to BBsi.left + ECW

[0015] Given these definitions, the following characteristics or features may be used by the inter-character break classifier component 106 to classify inter-character candidate breaks as a break point, a non-break point or an undecided point:
TABLE 1
f1 BBpi. width / ECH
f2 BBpi. height / ECH
f3 BBsi. width / ECH
f4 BBsi.height / ECH
f5 BBpi. top / ECH
f6 BBpi. bottom / ECH
f7 BBsi. top / ECH
f8 BBsi.bottom /ECH
f9 Bi.size / MBS
f10 Bi-1.size / MBS
f11 Bi+1.size / MBS
f12 (Bi.start - Bpi.start) / ECW
f13 Bpi.size / MBS
f14 (Bsi.start - Bi.start) / ECW
f15 Bsi.size / MBS


[0016] The last four features are used to establish the regularity of East Asian text. In a line that contains only East Asian characters, the character breaks are equidistant from one another. Thus, if there is a break at position x, another break can be expected to be located to the left of the current break at the position x - ECW and to the right of the current break at the position x + ECW. Thus, if there is a break at position x, adjacent breaks can be expected close to the positions x - ECW and x + ECW. Features f12 and f14 represent the measure or degree of this regularity between breaks. Breaks closest to the positions where they are expected are identified and expressed in terms of how well they match with their expected positions. Features f13 and f15 are included for completeness since they provide information about break size.

[0017] It should be noted that in those situations where symbols touching one another are divided in order to reproduce a missing break, the break size is equal to 1 (Bi.size = Bi.end - Bi.start + 1 = 1).

[0018] The distribution of the above features may be determined using training patterns to establish various combinations of feature values that characterize a break point, a non-break point and an undecided point. Once this has been determined, the values for these features may be used to classify inter-character candidate breaks for a textual line in an unknown image undergoing OCR. While the greatest accuracy will generally be achieved from an examination of all of these features, in some cases an examination of various subcombinations may be sufficient.

[0019] The proposed set of features presented above may be expanded with other similar geometric features to further improve classification accuracy. The expansion process may be accomplished by examining incidents of errors in classification to determine the scenarios when such errors occur.

[0020] The inter-character break classifier component 106 can employ any available classification technique (neural networks, decision trees, etc.) to perform the actual classification. Some classifiers will need training using the available feature set before they can be employed. A neural network such as a time delay neural network (TDNN) can be user to further improve the classification accuracy. In this approach, instead of simply using the aforementioned features to classify a particular candidate, the values for a set of features for a few (e.g., 1-3) preceding or successive break points may also be used in the classification process.

Character Recognition



[0021] Prior to sending the individual characters identified above to the character feature extraction component 112, some pre-processing is performed on the characters in order to improve accuracy. This step is illustrated in connection with FIGs. 6a and 6b. FIG. 6a shows the image of the original character along with the width and height of its bounding box. In the first step as shown in FIG. 6b, the original character is stretched by the character pre-processing component 108 so that in each dimension it extends to the edge or border of its bounding box, which now has a fixed dimension. In some cases the stretching step may include changing the aspect ratio of the character, in which case the the value of the original aspect ratio is retained for future use. One reason for performing the step illustrated in FIG6b is to remove inconsistencies between various features of the same character in different fonts.

[0022] After the characters have been properly stretched, the next step is to perform feature extraction with character feature extraction component 112. The features that are extracted may be any suitable set of features such as Gabor features or histogram features or any other feature set applicable to character recognition. If Gabor features are used, the feature extraction engine includes a bank of Gabor filters to identify repetitive structures that can be effectively characterized in the frequency domain. Since the OCR engine is based on a grayscale image, this means that instead of using pixel values of 0 or 1 at each pixel position, pixel values between 0 and 255 are used in the formulas for Gabor, histogram or any other features that may be employed. Additionally, a feature normalization step may be performed. For instance, in some examples features may be normalized by their maximum feature value. As previously mentioned, information concerning the original character aspect ratio may have been retained. This information may be added as an extra feature to the feature set that is employed.

[0023] After the aforementioned pre-processing step and feature computation step have been performed, character classification component 113 performs classification based on the input feature set received from the Character Feature Extraction component 112. The Character Classification Component 112 can employ any available classification technique (neural networks, decision trees, etc) to perform the actual classification. Some of these classifiers may require training using the available feature set before they can be employed. The character recognition engine 108 outputs a set of character guesses for each candidate character along with a probability for each guess.

Segmentation of Lines into Western and Hieroglyphic Text Segments



[0024] The process of dividing textual lines into Western and hieroglyphic text segments will be illustrated in connection with the textual line shown in FIGs. 7-12 and the flowchart shown in FIG. 13. First, in step 310, the inter-character break identifier component 104 and the inter-character break classifier component 106 perform the character breaking process described above to identify break points, non-break points and undecided points. Break points are indicated in FIG. 7 by lines B, non-break points by lines N and undecided points by lines U.

[0025] The Western and hieroglyphic text classifier component 114 in step 320 is used to identify the individual characters in the text, both Western and hieroglyphic characters. First, all non-break points are removed. Based on the confidence levels of the individual characters that are computed by the character recognition engine 108, the undecided break points are resolved. The resolution process may be performed by applying the well known Beam search algorithm, which will compute the most optimal sequence of characters between two consecutive break points. As can be seen in FIG. 8, undecided break points have been resolved and non-break points removed. It can be seen that the first undecided break point in FIG. 7 has been incorrectly removed in FIG. 8. At this point in time the sequence of characters has been identified for the given line and their confidence levels specified. In addition, the set of break points has been updated by removing some break points and confirming some others.

[0026] The Western and hieroglyphic text classifier component 114 continues the line segmentation process in step 330 by running any of a wide variety of well-known Western word breaking engines to obtain inter-word breaks in the given text line. The inter-word breaks are indicated by lines C in FIG. 9. During the course of this step some of the break points previously identified as inter-character break are now classified as inter-word breaks. From this point on, inter-word breaks will be used to denote the point at which different text segments are separated from one another.

[0027] In step 340 the Western and hieroglyphic text classifier component 114 places inter-word breaks around those characters that have been identified in step 320 as hieroglyphic characters with a level of confidence above some threshold. The threshold can be empirically determined. In some implementation inter-word breaks may be placed around some characters that are identified as hieroglyphic characters even if they have a confidence level below the threshold. This can be accomplished by examining some additional character features such as the character height relative to the height of the characters to its left and right, the identity of the characters to its left and right and the height of the character relative to the line height. In this way some additional characters can be identified as hieroglyphic characters, even though they initially had a confidence level below some predefined threshold. FIG. 10 shows inter-word breaks placed around the hieroglyphic characters that satisfy the above-mentioned conditions.

[0028] Next, in step 350 the Western and hieroglyphic text classifier component 114 counts the total number of characters N located between all consecutive inter-word breaks as well as the total number of pure Western characters W. If the ratio W / N is greater than some empirically determined threshold this text segment will be classified as a Western text segment, otherwise it will be classified as a hieroglyphic text segment. If a text segment is classified as a Western text segment any of a variety of well-known Western word recognizer engines will be invoked in step 360. If the confidence level for the recognized word provided by the Western word recognizer is lower than some threshold value the text segment will be re-classified as a hieroglyphic text segment. If the confidence level is above the threshold value the text segment will maintain its Western text segment classification. In Fig. 11 the words "Grand" and "Challenge" satisfy this condition and are recognized by the Western word recognizer with a confidence higher than a predefined threshold and, as a result, are classified as Western text segments. It should also be noted that the word "Grand" might be correctly recognized even though there is not break between the characters 'a' and 'n', which means that the Western word recognizer has its own logic for finding inter-character breaks and recognizing individual characters and words.. Word recognition results obtained from the word recognizer which are above the previously established threshold are preserved as the final recognition results for the corresponding text segment and later passed to the output component 116.

[0029] For the remaining text segments, which are presumably all hieroglyphic text segments, a final segmentation process is performed in step 370 by denoting all breaks around any remaining hieroglyphic characters as inter-word breaks. This step is illustrated in FIG. 12. In this way characters that are between two consecutive inter-word breaks represent a single word, with a string of characters representing Western words in the case of Western text segments and single character words in the case of hieroglyphic text segments.

[0030] All recognition results along with the position of inter-word breaks are passed to the output component 116. In case of Western text segments, the results of the Western word recognition process are passed to the output component, while in case of a hieroglyphic text segments the results of the character recognition engine 108 that were resolved in step 320 are passed to the output component.

[0031] As used in this application, the terms "component," "module," "engine," "system," "apparatus," "interface," or the like are generally intended to refer to a computer-related entity, either hardware, a combination of hardware and software, software, or software in execution. For example, a component may be, but is not limited to being, a process running on a processor, a processor, an object, an executable, a thread of execution, a program, and/or a computer. By way of illustration, both an application running on a controller and the controller can be a component. One or more components may reside within a process and/or thread of execution and a component may be localized on one computer and/or distributed between two or more computers.

[0032] Furthermore, the claimed subject matter may be implemented as a method, apparatus, or article of manufacture using standard programming and/or engineering techniques to produce software, firmware, hardware, or any combination thereof to control a computer to implement the disclosed subject matter. The term "article of manufacture" as used herein is intended to encompass a computer program accessible from any computer-readable device, carrier, or media. For example, computer readable media can include but are not limited to magnetic storage devices (e.g., hard disk, floppy disk, magnetic strips ...), optical disks (e.g., compact disk (CD), digital versatile disk (DVD) . ..), smart cards, and flash memory devices (e.g., card, stick, key drive ...).


Claims

1. A method for performing character recognition on an input image, comprising:

receiving an input image that includes at least one textual line including multiple characters;

identifying candidate inter-character breaks along the textual line;

classifying each of the candidate inter-character breaks as an actual break, a non-break for being removed in subsequent processing or an indeterminate break for being resolved in subsequent processing using at least in part combinations of geometrical properties of each respective candidate inter-character break and character bounding boxes adjacent thereto; and

recognizing the candidate characters in the textual line based at least in part on a feature set extracted from each respective candidate character, and

wherein the intermediate break is resolved based on the confidence levels of the recognized candidate characters in the character bounding boxes adjacent thereto.


 
2. The method of claim 1 wherein the geometrical properties on which each of the candidate inter-character breaks are classified include a size and location of the adjacent character bounding boxes relative to an estimated character height for the textual line.
 
3. The method of claim 2 wherein the geometrical properties on which each of the candidate inter-character breaks are classified further include a size of the respective candidate inter-character break and its neighboring breaks relative to an average break size for the textual line.
 
4. The method of claim 3 wherein the geometrical properties on which each of the candidate inter-character breaks are classified further include a degree of regularity in estimated character widths for the candidate characters in the textual line.
 
5. The method of claim 1 wherein recognizing the candidate characters includes stretching the candidate characters on an input gray scale image.
 
6. The method of claim 1 wherein recognizing the candidate characters includes performing feature computation using a character feature set and performing classification based on the character feature set by using any current or future classification algorithm.
 
7. An image processing apparatus for performing character recognition on an input image, comprising:

an input component that receives an input image that includes at least one textual line including multiple characters;

an inter-character break identifier component that identifies candidate inter- character breaks along the textual line;

an inter-character break classifier component that classifies each of the candidate inter-character breaks as an actual break, a non-break for being removed in subsequent processing or an indeterminate break for being resolved in subsequent processing using at least in part combinations of geometrical properties of each respective candidate inter-character break and character bounding boxes adjacent thereto;

a character recognition component that recognizes the candidate characters in the textual line based at least in part on a feature set extracted from each respective candidate character; and

wherein the inter-character break classifier component is further configured to resolve the intermediate break based on the confidence levels of the recognized candidate characters in the character bounding boxes adjacent thereto.


 
8. The image processing apparatus of claim 7, wherein the at least one textual line includes Western characters and hieroglyphic characters and wherein the image processing apparatus is further configured to segment the Western and hieroglyphic characters of the at least one textual line,
the image processing apparatus further comprising,
a Western and hieroglyphic text classifier component that segments the at least one textual line into Western text segment and hieroglyphic text segment; and
an output component that receives line segmentation and recognition results.
 
9. The image processing apparatus of claim 8 wherein the Western and hieroglyphic text classifier component is configured to remove non-break points and resolve undecided points and identify individual Western and Hieroglyphic characters.
 
10. The image processing apparatus of claim 9 wherein the Western and hieroglyphic text classifier component is further configured to identify inter-word breaks in the textual line based on a Western word breaking algorithm.
 
11. The image processing apparatus of claim 10 wherein the Western and hieroglyphic text classifier component is configured to place inter- word breaks around each character that has been recognized as a hieroglyphic character with a confidence level greater than a threshold level.
 
12. The image processing apparatus of claim 11 wherein the Western and hieroglyphic text classifier component is configured to place inter- word breaks around each additional character that has been recognized as a hieroglyphic character with a confidence level lower than a threshold level by examining at least one additional character feature.
 
13. The image processing apparatus of claim 12 wherein the additional character features include character height relative to a height of characters to the left and right of the character, an identity of characters to the left and right of the character and a height of the character relative to line height.
 
14. The image processing apparatus of claim 13 wherein the Western and hieroglyphic text classifier component is configured to determine a ratio of a number of Western characters in a text segment located between consecutive inter- word breaks and a total number of characters in the text segment and classifying the text segment as a Western text segment if the ratio exceeds a predetermined threshold or an hieroglyphic text segment otherwise.
 
15. The image processing apparatus of claim 14 further comprising a Western word recognizer engine for recognizing words in the Western text segment.
 
16. The image processing apparatus of claim 15 wherein the Western word recognizer engine provides a Western word recognition result and a confidence level associated therewith, wherein the confidence level represents a probability that recognized words have been recognized correctly and wherein the Western and hieroglyphic text classifier component is further configured to reclassify the Western text segment as a hieroglyphic text segment if the confidence level is below a threshold level.
 


Ansprüche

1. Verfahren zum Durchführen von Zeichenerkennung mit einem Eingabebild, umfassend:

Empfangen eines Eingabebilds, das mindestens eine Textlinie mit mehreren Zeichen beinhaltet;

Identifizieren von möglichen Zwischenräumen zwischen Zeichen entlang der Textlinie; Klassifizieren jedes der möglichen Zwischenräume zwischen Zeichen als einen tatsächlichen Zwischenraum, einen Nichtzwischenraum zum Entfernen während des anschließenden Verarbeitens oder einen unbestimmten Zwischenraum zum Klären während des anschließenden Verarbeitens unter Verwendung von mindestens teilweise Kombinationen von geometrischen Eigenschaften jedes jeweiligen möglichen Zwischenraums zwischen Zeichen und daran angrenzenden zeichenbegrenzenden Feldern; und

Erkennen der möglichen Zeichen in der Textlinie auf der Grundlage von mindestens teilweise einem Merkmalsatz, der aus jedem der jeweiligen möglichen Zeichen extrahiert wurde, und

wobei der unbestimmte Zwischenraum auf der Grundlage von Konfidenzniveaus der erkannten möglichen Zeichen in den daran angrenzenden zeichenbegrenzenden Feldern geklärt wird.


 
2. Verfahren nach Anspruch 1, wobei die geometrischen Eigenschaften, gemäß denen die möglichen Zwischenräume zwischen Zeichen klassifiziert werden, eine Größe und einen Ort der angrenzenden zeichenbegrenzenden Felder in Bezug auf eine geschätzte Zeichenhöhe der Textlinie beinhalten.
 
3. Verfahren nach Anspruch 2, wobei die geometrischen Eigenschaften, gemäß denen die möglichen Zwischenräume zwischen Zeichen klassifiziert werden, ferner eine Größe des jeweiligen möglichen Zwischenraums zwischen Zeichen und dessen angrenzenden Zwischenräumen in Bezug auf eine durchschnittliche Zwischenraumgröße der Textlinie beinhalten.
 
4. Verfahren nach Anspruch 3, wobei die geometrischen Eigenschaften, gemäß denen die möglichen Zwischenräume zwischen Zeichen klassifiziert werden, ferner einen Regelmäßigkeitsgrad der geschätzten Zeichenbreiten der möglichen Zeichen in der Textlinie beinhalten.
 
5. Verfahren nach Anspruch 1, wobei das Erkennen der möglichen Zeichen das Dehnen der möglichen Zeichen auf einem Grauskala-Eingabebild beinhaltet.
 
6. Verfahren nach Anspruch 1, wobei das Erkennen der möglichen Zeichen das Durchführen einer Merkmalberechnung unter Verwendung eines Zeichenmerkmalsatzes und das Durchführen einer Klassifizierung auf der Grundlage des Zeichenmerkmalsatzes unter Verwendung jedes derzeitigen oder künftigen Klassifizierungsalgorithmus beinhaltet.
 
7. Bildverarbeitungsvorrichtung zum Durchführen von Zeichenerkennung mit einem Eingabebild, umfassend:

eine Eingabekomponente zum Empfangen eines Eingabebilds, das mindestens eine Textlinie mit mehreren Zeichen beinhaltet;

eine Identifizierungskomponente für Zwischenräume zwischen Zeichen zum Identifizieren von möglichen Zwischenräumen zwischen Zeichen entlang der Textlinie;

eine Klassifizierungskomponente für Zwischenräume zwischen Zeichen zum Klassifizieren jedes der möglichen Zwischenräume zwischen Zeichen als einen tatsächlichen Zwischenraum, einen Nichtzwischenraum zum Entfernen während des anschließenden Verarbeitens oder einen unbestimmten Zwischenraum zum Klären während des anschließenden Verarbeitens unter Verwendung von mindestens teilweise Kombinationen von geometrischen Eigenschaften jedes jeweiligen möglichen Zwischenraums zwischen Zeichen und daran angrenzenden zeichenbegrenzenden Feldern;

eine Zeichenerkennungskomponente zum Erkennen der möglichen Zeichen in der Textlinie auf der Grundlage von mindestens teilweise einem Merkmalsatz, der aus jedem der jeweiligen möglichen Zeichen extrahiert wurde; und

wobei die Klassifizierungskomponente für Zwischenräume zwischen Zeichen weiterhin zum Klären des unbestimmten Zwischenraums auf der Grundlage von Konfidenzniveaus der erkannten möglichen Zeichen in den daran angrenzenden zeichenbegrenzenden Feldern konfiguriert ist.


 
8. Bildverarbeitungsvorrichtung nach Anspruch 7, wobei die mindestens eine Textzeile lateinische Zeichen und Hieroglyphenzeichen beinhaltet und wobei die Bildverarbeitungsvorrichtung weiterhin zum Segmentieren der lateinischen und der Hieroglyphenzeichen der mindestens einen Textzeile konfiguriert ist,
wobei die Bildverarbeitungsvorrichtung ferner umfasst
eine Klassifizierungskomponente für Lateinzeichen- und Hieroglyphentext, die die mindestens eine Textzeile in ein Lateinzeichen-Textsegment und ein Hieroglyphentextsegment segmentiert; und
eine Ausgabekomponente, die die Zeilensegmentierung und die Erkennungsergebnisse empfängt.
 
9. Bildverarbeitungsvorrichtung nach Anspruch 8, wobei die Klassifizierungskomponente für Lateinzeichen- und Hieroglyphentext zum Entfernen von Nichtzwischenraumpunkten und zum Klären unbestimmter Punkte und zum Identifizieren einzelner lateinischer und Hieroglyphenzeichen konfiguriert ist.
 
10. Bildverarbeitungsvorrichtung nach Anspruch 9, wobei die Klassifizierungskomponente für Lateinzeichen- und Hieroglyphentext ferner zum Identifizieren von Zwischenräumen zwischen Wörtern in der Textzeile auf der Grundlage eines Algorithmus für Zwischenräume zwischen Lateinzeichenwörtern konfiguriert ist.
 
11. Bildverarbeitungsvorrichtung nach Anspruch 10, wobei die Klassifizierungskomponente für Lateinzeichen- und Hieroglyphentext zum Setzen von Zwischenräumen zwischen Wörtern um jedes Zeichen konfiguriert ist, das mit einem Konfidenzniveau, das größer als ein Schwellenniveau ist, als Hieroglyphenzeichen identifiziert wurde.
 
12. Bildverarbeitungsvorrichtung nach Anspruch 11, wobei die Klassifizierungskomponente für Lateinzeichen- und Hieroglyphentext zum Setzen von Zwischenräumen zwischen Wörtern um jedes zusätzliche Zeichen konfiguriert ist, das mit einem Konfidenzniveau, das kleiner als ein Schwellenniveau ist, durch Untersuchen mindestens eines zusätzlichen Zeichenmerkmals als Hieroglyphenzeichen identifiziert wurde.
 
13. Bildverarbeitungsvorrichtung nach Anspruch 12, wobei die zusätzlichen Zeichenmerkmale eine Zeichenhöhe in Bezug auf eine Höhe von Zeichen links und rechts vom Zeichen, eine Identität von Zeichen links und rechts vom Zeichen und eine Höhe des Zeichens in Bezug auf eine Linienhöhe beinhalten.
 
14. Bildverarbeitungsvorrichtung nach Anspruch 13, wobei die Klassifizierungskomponente für Lateinzeichen- und Hieroglyphentext zum Bestimmen eines Verhältnisses der Anzahl von lateinischen Zeichen in einem Textsegment, die sich zwischen aufeinanderfolgenden Zwischenräumen zwischen Wörtern befinden, und einer Gesamtanzahl von Zeichen im Textsegment und zum Klassifizieren des Textsegments als Lateinzeichen-Textsegment, wenn das Verhältnis eine vorbestimmte Schwelle übersteigt, oder anderenfalls als Hieroglyphentextsegment konfiguriert ist.
 
15. Bildverarbeitungsvorrichtung nach Anspruch 14, ferner umfassend eine Erkennungsmaschine für Lateinzeichenwörter zum Erkennen von Wörtern im Lateinzeichen-Textsegment.
 
16. Bildverarbeitungsvorrichtung ach Anspruch 15, wobei die Erkennungsmaschine für Lateinzeichenwörter ein Erkennungsergebnis für ein Lateinzeichenwort und ein damit assoziiertes Konfidenzniveau bereitstellt, wobei das Konfidenzniveau eine Wahrscheinlichkeit repräsentiert, dass die erkannten Wörter korrekt erkannt wurden, und wobei die Klassifizierungskomponente für Lateinzeichen- und Hieroglyphentext ferner zum Neuklassifizieren des Lateinzeichen-Textsegments als Hieroglyphentextsegment konfiguriert ist, wenn das Konfidenzniveau unter einem Schwellenniveau liegt.
 


Revendications

1. Procédé de mise en œuvre de reconnaissance de caractères sur une image d'entrée, comprenant les étapes ci-dessous consistant à :

recevoir une image d'entrée qui inclut au moins une ligne textuelle incluant de multiples caractères ;

identifier des ruptures entre caractères candidates le long de la ligne textuelle ;

classer chacune des ruptures entre caractères candidates en tant qu'une rupture réelle, qu'une non-rupture devant être supprimée dans le cadre d'un traitement subséquent, ou qu'une rupture indéterminée devant être résolue dans le cadre d'un traitement subséquent, en utilisant, au moins en partie, des combinaisons de propriétés géométriques de chaque rupture entre caractères candidate respective et des rectangles englobants de caractères adjacents à celle-ci ; et

reconnaître les caractères candidats dans la ligne textuelle, sur la base, au moins en partie, d'un ensemble de caractéristiques extrait de chaque caractère candidat respectif ; et

dans lequel la rupture indéterminée est résolue sur la base des niveaux de confiance des caractères candidats reconnus dans les rectangles englobants de caractères adjacents à celle-ci.


 
2. Procédé selon la revendication 1, dans lequel les propriétés géométriques sur lesquelles chacune des ruptures entre caractères candidates est classée incluent une taille et un emplacement des rectangles englobants de caractères adjacents par rapport à une hauteur de caractères estimée pour la ligne textuelle.
 
3. Procédé selon la revendication 2, dans lequel les propriétés géométriques sur lesquelles chacune des ruptures entre caractères candidates est classée incluent en outre une taille de la rupture entre caractères candidate respective et de ses ruptures voisines par rapport à une taille de rupture moyenne pour la ligne textuelle.
 
4. Procédé selon la revendication 3, dans lequel les propriétés géométriques sur lesquelles chacune des ruptures entre caractères candidates est classée incluent en outre un degré de régularité dans des largeurs de caractères estimées pour les caractères candidats dans la ligne textuelle.
 
5. Procédé selon la revendication 1, dans lequel l'étape de reconnaissance des caractères candidats inclut l'étape consistant à étirer les caractères candidats sur une image à niveaux de gris d'entrée.
 
6. Procédé selon la revendication 1, dans lequel l'étape de reconnaissance des caractères candidats inclut l'étape consistant à mettre en œuvre un calcul de caractéristiques en utilisant un ensemble de caractéristiques de caractères, et l'étape consistant à mettre en œuvre une classification sur la base de l'ensemble de caractéristiques de caractères, en utilisant tout algorithme de classification actuel ou futur.
 
7. Appareil de traitement d'image destiné à mettre en œuvre une reconnaissance de caractères sur une image d'entrée, comprenant :

un composant d'entrée qui reçoit une image d'entrée qui inclut au moins une ligne textuelle incluant de multiples caractères ;

un composant d'identification de ruptures entre caractères qui identifie des ruptures entre caractères candidates le long de la ligne textuelle ;

un composant de classification de ruptures entre caractères qui classe chacune des ruptures entre caractères candidates en tant qu'une rupture réelle, qu'une non-rupture devant être supprimée dans le cadre d'un traitement subséquent ou qu'une rupture indéterminée devant être résolue dans le cadre d'un traitement subséquent en utilisant, au moins en partie, des combinaisons de propriétés géométriques de chaque rupture entre caractères candidate respective et des rectangles englobants de caractères adjacents à celle-ci ;

un composant de reconnaissance de caractères qui reconnaît les caractères candidats dans la ligne textuelle, sur la base, au moins en partie, d'un ensemble de caractéristiques extrait de chaque caractère candidat respectif ; et

dans lequel le composant de classification de ruptures entre caractères est en outre configuré de manière à résoudre la rupture indéterminée sur la base des niveaux de confiance des caractères candidats reconnus dans les rectangles englobants de caractères adjacents à celle-ci.


 
8. Appareil de traitement d'image selon la revendication 7, dans lequel ladite au moins une ligne textuelle inclut des caractères occidentaux et des caractères hiéroglyphiques, et dans lequel l'appareil de traitement d'image est en outre configuré de manière à segmenter les caractères occidentaux et hiéroglyphiques de ladite au moins une ligne textuelle ;
l'appareil de traitement d'image comprenant en outre :

un composant de classification de texte occidental et hiéroglyphique qui segmente ladite au moins une ligne textuelle en segment de texte occidental et segment de texte hiéroglyphique ; et

un composant de sortie qui reçoit des résultats de reconnaissance et de segmentation de lignes.


 
9. Appareil de traitement d'image selon la revendication 8, dans lequel le composant de classification de texte occidental et hiéroglyphique est configuré de manière à supprimer des points sans rupture, à résoudre des points indéterminés, et à identifier des caractères occidentaux et hiéroglyphiques individuels.
 
10. Appareil de traitement d'image selon la revendication 9, dans lequel le composant de classification de texte occidental et hiéroglyphique est en outre configuré de manière à identifier des ruptures entre mots dans la ligne textuelle sur la base d'un algorithme de rupture de mots occidentaux.
 
11. Appareil de traitement d'image selon la revendication 10, dans lequel le composant de classification de texte occidental et hiéroglyphique est configuré de manière à placer des ruptures entre mots autour de chaque caractère qui a été reconnu comme étant un caractère hiéroglyphique avec un niveau de confiance supérieur à un niveau de seuil.
 
12. Appareil de traitement d'image selon la revendication 11, dans lequel le composant de classification de texte occidental et hiéroglyphique est configuré de manière à placer des ruptures entre mots autour de chaque caractère supplémentaire qui a été reconnu comme étant un caractère hiéroglyphique avec un niveau de confiance inférieur à un niveau de seuil, en examinant au moins une caractéristique de caractère supplémentaire.
 
13. Appareil de traitement d'image selon la revendication 12, dans lequel les caractéristiques de caractères supplémentaires incluent une hauteur de caractère par rapport à une hauteur de caractères à gauche et à droite du caractère, une identité de caractères à gauche et à droite du caractère et une hauteur du caractère par rapport à une hauteur de ligne.
 
14. Appareil de traitement d'image selon la revendication 13, dans lequel le composant de classification de texte occidental et hiéroglyphique est configuré de manière à déterminer un rapport d'un nombre de caractères occidentaux dans un segment de texte situé entre des ruptures entre mots consécutives et d'un nombre total de caractères dans le segment de texte, et à classer le segment de texte comme un segment de texte occidental si le rapport dépasse un seuil prédéterminé, ou sinon comme un segment de texte hiéroglyphique.
 
15. Appareil de traitement d'image selon la revendication 14, comprenant en outre un moteur de reconnaissance de mots occidentaux destiné à reconnaître des mots dans le segment de texte occidental.
 
16. Appareil de traitement d'image selon la revendication 15, dans lequel le moteur de reconnaissance de mots occidentaux fournit un résultat de reconnaissance de mots occidentaux et un niveau de confiance associé à celui-ci, dans lequel le niveau de confiance représente une probabilité que des mots reconnus aient été reconnus correctement, et dans lequel le composant de classification de texte occidental et hiéroglyphique est en outre configuré de manière à reclasser le segment de texte occidental comme un segment de texte hiéroglyphique si le niveau de confiance est inférieur à un niveau de seuil.
 




Drawing




















Cited references

REFERENCES CITED IN THE DESCRIPTION



This list of references cited by the applicant is for the reader's convenience only. It does not form part of the European patent document. Even though great care has been taken in compiling the references, errors or omissions cannot be excluded and the EPO disclaims all liability in this regard.

Patent documents cited in the description