WO2009070032A1 - A method for processing optical character recognition (ocr) data, wherein the output comprises visually impaired character images - Google Patents
A method for processing optical character recognition (ocr) data, wherein the output comprises visually impaired character images Download PDFInfo
- Publication number
- WO2009070032A1 WO2009070032A1 PCT/NO2008/000419 NO2008000419W WO2009070032A1 WO 2009070032 A1 WO2009070032 A1 WO 2009070032A1 NO 2008000419 W NO2008000419 W NO 2008000419W WO 2009070032 A1 WO2009070032 A1 WO 2009070032A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- images
- template
- character
- image
- visually impaired
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/148—Segmentation of character regions
- G06V30/155—Removing patterns interfering with the pattern to be recognised, such as ruled lines or underlines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/28—Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/19—Recognition using electronic means
- G06V30/191—Design or setup of recognition systems or techniques; Extraction of features in feature space; Clustering techniques; Blind source separation
- G06V30/1914—Determining representative reference patterns, e.g. averaging or distorting patterns; Generating dictionaries, e.g. user dictionaries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/01—Solutions for problems related to non-uniform document background
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
Definitions
- OCR Optical Character Recognition
- the present invention is related to a method for resolving contradicting output data from an Optical Character Recognition (OCR) system, and especially to a method for processing OCR output data, wherein the output data comprises unrecognizable character images due to crossed out text or other types of overlaid objects impairing the optical recognition process of the OCR system
- OCR Optical Character Recognition
- Optical character recognition systems provide a transformation of pixelized images of documents into ASCII coded text which facilitates searching, substitution, reformatting of documents etc. in a computer system.
- One aspect of OCR functionality is to convert handwritten and typewriter typed documents, books, medical journals, etc. into for example Internet or Intranet searchable documents.
- the quality of inforaiation retrieval and document searching is considerably enhanced if all documents are electronically retrievable and searchable.
- a company Intranet system can link together all old and new documents of an enterprise through extensive use of OCR functionality implemented as a part of the Intranet (or as part of the Internet if the documents are of public interest).
- the quality of the OCR functionality is limited due to the fact that the complexity of an OCR system in itself is a challenge. It is difficult to provide an OCR functionality that can solve any problem encountered when trying to convert images of text into computer coded text.
- One such problem is due to crossed out text that often may be encountered in documents. For example, a stamp with the text "COPY" may be applied onto a page of a document to signify that this document is not the original document, but a copy of the original document.
- documents have to be certified as a correct copy of the original document, which is typically done with additional stamps and a signature of a person entrusted to certify such copies, for example.
- OCR systems provide output data comprising a list of uncertainly recognized characters. Such crossed out characters etc. will therefore be identifiable as such, and their position on a text page, in words etc., possible alternative interpretations of the hidden or partly hidden character etc. may be reported by the OCR system.
- such overlaid objects may be identified, 5 as such in it self, but also the location and extent of the overlaid object across an image of a character in a text may be identified.
- the visually impaired parts of the character image is discarded from the image representing the character.
- the remains of the visually impaired character images are then compared with images from a set of template character images according to the present invention.o This comparison will then make it possible to identify at least one template image having a highest similarity with the image comprising the remains of the visually impaired character image. If this process returns only one candidate template image having a similarity above a preset threshold level, the problem would have been solved.
- the problem is to select the correct candidate character image amongst a plurality of candidate template images as the correct identification of the visually impaired character.
- the information related to location and extent of the overlaid objects of a character is used to model an overlaid object at a same location and extent in template images before comparing the visually impaired character instance with the template images modelled with the actual overlaid object.
- correlations between the remaining parts of images of thes respective visually impaired character instance and possible candidate template images provides a method for identifying which template character image that actually represents the visually impaired character instance, and thereby the correct recognition of the visually impaired character instance.
- the possibility to select the correct candidate character image amongst a plurality of candidate characters as the correct identification of the visually impaired character is significantly improved when a set of template character images is provided for by using images of characters from the document itself identified by the OCR system as being characters identified with aS certainty above a predefined threshold level.
- Such character images comprises details of the character images as encountered in the document itself, and not as modelled character images as known in prior art for character template matching.
- Figure 1 illustrates an example of the stamped letter P visually impairing a part of a text 5 page.
- the character P is from a stamp comprising the text "COPY".
- Figure 2a illustrates a visually impaired character instance of a character 'a' partly being overlaid by the P in figure 1.
- Figure 2b illustrates removal of the visually impaired section in figure 2b according to the present invention.
- Figure 3 a illustrates an example of maximum correlation between the image in figure 2b and a template for the character 'a', as depicted in figure 3b.
- Figure 4a illustrates different categories of pixels in the visually impaired character instance in figure 2a, according to the present invention.
- Figure 4b illustrates corresponding categories of pixels in a crossed out template 20 according to the present invention.
- the presence of an overlaid object for example due to a stamp or handwritten signature, etc. may be identified by analyzing connected pixels in images comprising text rich regions in a document.
- an overlaid object for example due to a stamp or handwritten signature, etc.
- the output data from an OCR system may comprise identifications of uncertainly recognized characters, their positions on a page etc. providing a hint about where such problems may be present.
- a hint of such a problem can be identified by the fact that where a text line is crossed out there will be at least one character that cannot be recognized, and other characters of the same poor quality will be identified.
- the identification of crossed out text will then trigger a search for the actual images of characters comprising crossed out regions. These characters will be identified as unrecognizable characters or as uncertainly recognized characters by the OCR system. For example, by comparing such reported characters and their position in the image of5 the text with information obtained when identifying the presence of the overlaying imprint, these characters are identified. Such character images are then being identified by providing a bounding box around the respective characters in the text. These character instances will then comprise three types of possible crossed out character images: one type comprising a completely crossed out image, and a second typeQ comprising a partly crossed out character image, wherein the crossed out section covers parts of the body of the character.
- the third possible scenario is that the crossed out section in the bounding box do not cross or touch the body of the character itself, but passes the bounding box without covering any part of the body of the character.
- the OCR system is incapable to identify5 correctly the character since the extra "on" pixels will probably degrade the recognition algorithm used by the OCR system.
- the characters being completely crossed out can be identified by using the words they are part of for their certain identification.
- a word comprising a missing character (or completely unrecognizable character) would provide a list of possible candidate words.
- the overlaid objectQ masks several characters in the words. The resolve of such contradicting words represent another type of invention which is outside the scope of the present invention.
- the present invention will provide certain recognition of partial crossed out character images (the type two and three referenced above).
- One aspect of the present invention is to be able to reference parts or locations of regions in images across different images of characters.
- a common reference system may be established by using for example a corner of a text page as a referencing point, and then using ordered pairs of numbers to reference individual pixels, cluster of pixels, connected pixels etc.
- the location or region information about differentiating aspects of character images and/or uncertainly recognized characters can be referenced consistent throughout all the images of the individual respective images of the characters when the images are aligned with respect to each other such that as much as possible of the character bodies themselves are positioned on top of each other when the images are viewed as being placed on top of each other.
- displacements between the different images can be calculated, for example by correlating combinations of images.
- the bonding boxes used to separate the crossed out characters from the text page can be correlated with template images of characters thereby providing a displacement used for an alignment of the images according to the present invention.
- the alignment provided by calculating maximum correlation and the corresponding placement between images makes it possible to identify the regions of visually impaired character instances in an image of a text page that are actually crossed out relative to the character body itself.
- the region is identified, the same region comprising the crossed out parts can be introduced at the same location relative to the character body itself in template images of characters. This provides then a possibility to examine what is left of the template images, and compare the content of these regions with the regions of the visually impaired character instance comprising the visually impaired sections.
- a search for maximum correlation provides a measurement of which template image that comprises regions outside the visually impaired sections that is most equal the same regions in the visually impaired character instance comprising the overlaid object.
- the template image having the maximum correlation value is the correct selection as an identification of the visually impaired crossed out character instance.
- Figure 1 illustrates a situation wherein a letter P from the word "COPY" that has been stamped across a section of a page.
- the crosses out character 'a' indicated by the arrow in fig. 1 will be used to illustrate examples aspects of the present invention.
- Figure 2a illustrates the character 'a' provided with a bounding box separating the character 'a' image comprising the visually impaired section out from the rest of the image comprising the text.
- Figure 2b illustrates the situation wherein the region comprising the visually impaired section has been "blanked out", by for example setting all grey level pixel values to the background colour (i.e. white).
- comparing the image when the crossed out section is "blanked out” simplifies the comparison with other character images.
- the image in fig. 2b is then used when comparing the image with other template images.
- an imprint for example from a stamp
- information about the points where the imprint crosses text lines can be used to identify the entry and exit points of the imprint in the bonding box of the visually impaired character instance, since the visually impaired character instance is actually fetched from the position on the page wherein the imprint actually is present.
- any method providing an identification of the regions comprising the overlaid object is within the scope of the present invention.
- the size of the imprint can be used to determine if the visually impaired character image is of type one as detailed above, for example. If the type is the type one, the process is terminated, the other two types of overlaid objects are processed the same way according to the present invention.
- the image in figure 2b is correlated with template images created from images of characters having an image quality above a predefined threshold value.
- Figure 3 a illustrates how the correlation identified by correlating the image in figure 2b with the template image in figure 3b provides an alignment between the visually impaired image of the character 'a' inside the chosen bonding box, and how it is possible to identify which regions belongs to the character itself, and which regions belongs to the crossed out regions.
- character body sizes of template images are compared with the size of the extent of the overlaid object in the visually impaired character instance. If the size of the character body of a template image (scaled to the font size of the document) is below the size of the extent of the visually impaired section, these template images are discarded from further use when searching for the correct identification of the visually impaired character instance, since these template image bodies would be completely hidden by for example the imprint visually impairing the character images. In this manner, the number of comparisons can be limited.
- Figure 4a and figure 4b illustrates how pixels in the images representing the template, s the visually impaired instance and the crossed out sections can be classified as belonging to one of four different classifications.
- This classification is best achieved by thresholding the template image providing a highest correlation with the visually impaired template instance comprising the crossed out section, then introducing the identified crossed out section into this template image. Since the correlation between theQ images provides the displacement or offset between the images, these classifications can be identified in all the images as represented in figure 4a and 4b, respectively, and can be visualized by letting:
- 'Off pixels touched by the crossed out section be illustrates as black 'x'.
- Q There are several methods for investigating which template image that has a maximum similarity when compared with the visually impaired character image. For example by correlation of images after alignment of the respective images.
- the only interesting pixels to investigate are the 'on' pixels as identified above in the classification of pixels 5 comprised in the images being processed in the OCR system. Then it is possible to calculate a measure of parallelism between the 'on' pixels in the visually impaired character instance and the template image.
- An example of such parallelism is:
- p k are the offset untouched 'on' pixels in the visually impaired character instance and p ⁇ the untouched 'on'-pixels in the template image.
- the parallelism for the 'a' template example from figure 1 is 0.971 which is the highest of all templates identified in this document, and a correct classification is achieved.
- the method steps described above is 5 best achieved when the template images are images of characters as they actually appears in the document being processed by the OCR system.
- the template images can be identified as character images of a quality above a predefined level and/or as a super positioning of several images on top of each other representing the same character image, denoted as a character class. 0
- the accumulation of aligned images into the template for a class representing a character comprises further adding corresponding grey level pixel values from corresponding locations in the aligned images together such that each grey level pixel value iso weighted with an inverse of the number of currently accumulated aligned images in the template image for the class before performing the addition.
- a typical confusion alternative for the character c is the character e. It is reasonable to expect that there will be a template present for the character e, but probably not for the character c.
- a synthetic template image can be provided for on basis of an already existing template image that resembles the missing template image. In this way, some of the graphical attributes of characters as they are encountered in the document being processed by the OCR system will be part of the synthesized templates image.
- an occluded character is recognized through performing following method steps, for example embodied in a computer program in communication or linked to an OCR system in the computer system, or to another computer system via a network:
- a bounding box for the occluted character/word is identified, and there is also identified a location measure of the clutter relative to the bounding box.
- Step 1 set the clutter in the character image to a representative background level (black inverted to white, or vice versa for example),
- Step 2 correltae with accumulated templates
- Step 3 threshold accumulated templates, for example by Canny thresholding,
- Step 4 for each template candidate:
- the template with the maximum value is the most probable character for the occluded character, and thereby the occluded character is identified.
- Embodiments 1.
- a method for resolving contradicting output data from an Optical Character Recognition (OCR) system wherein the output data comprises at least one visually impaired character instance due to at least one overlaid object in a document being processed by the OCR system, the method comprises:
- p' k are pixel values from the image of the at least one visually impaired character instance belonging to the visually impaired character instance body itself, and not the background of the image
- p k is correspondingly located pixel values from a respective aligned image of a template image belonging to the template character body itself, and not the background of the image.
- step providing a template set comprises sorting all reported identified characters above the threshold level into classes, wherein each class represents the same identified character in the template set, and then performing the steps of providing images for each character or class in the template set by:
- IV. for the initial pair identified in I or II correlate the images of the pair identifying a displacement between them before generating an aligned accumulated template image of the images that are used as a starting template image for the respective character class, V. for all the remaining samples in the class, correlate them with the accumulated template in IV to identify a displacement between them, and if the correlation is above the predefined threshold, align the images before adding the aligned images to the accumulated image in IV, VI. if some of the aligned images used in V comprise parts of the image that is outside the starting template in IV, expand the accumulated template after all samples have been used.
- the accumulation of aligned images into the template for a class representing a character further comprises adding corresponding grey level pixel values from corresponding locations in the alignedo images together such that each grey level pixel value is weighted with an inverse of the number of currently accumulated aligned images in the template image for the class before performing the addition.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Character Input (AREA)
- Character Discrimination (AREA)
- Image Processing (AREA)
- Editing Of Facsimile Originals (AREA)
Abstract
Description
Claims
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US12/744,974 US8467614B2 (en) | 2007-11-28 | 2008-11-21 | Method for processing optical character recognition (OCR) data, wherein the output comprises visually impaired character images |
EP08855111A EP2220590A1 (en) | 2007-11-28 | 2008-11-21 | A method for processing optical character recognition (ocr) data, wherein the output comprises visually impaired character images |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
NO20076153 | 2007-11-28 | ||
NO20076153 | 2007-11-28 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2009070032A1 true WO2009070032A1 (en) | 2009-06-04 |
Family
ID=40678786
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/NO2008/000419 WO2009070032A1 (en) | 2007-11-28 | 2008-11-21 | A method for processing optical character recognition (ocr) data, wherein the output comprises visually impaired character images |
Country Status (4)
Country | Link |
---|---|
US (1) | US8467614B2 (en) |
EP (1) | EP2220590A1 (en) |
RU (1) | RU2445699C1 (en) |
WO (1) | WO2009070032A1 (en) |
Families Citing this family (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2010157107A (en) * | 2008-12-26 | 2010-07-15 | Hitachi Software Eng Co Ltd | Business document processor |
US8792748B2 (en) * | 2010-10-12 | 2014-07-29 | International Business Machines Corporation | Deconvolution of digital images |
US20120212606A1 (en) * | 2011-02-20 | 2012-08-23 | Min-Hung Chien | Image processing method and image processing apparatus for dealing with pictures found by location information and angle information |
JP5701181B2 (en) * | 2011-08-18 | 2015-04-15 | 株式会社Pfu | Image processing apparatus, image processing method, and computer program |
JP5701182B2 (en) | 2011-08-18 | 2015-04-15 | 株式会社Pfu | Image processing apparatus, image processing method, and computer program |
US9058644B2 (en) * | 2013-03-13 | 2015-06-16 | Amazon Technologies, Inc. | Local image enhancement for text recognition |
WO2014204337A1 (en) * | 2013-06-18 | 2014-12-24 | Abbyy Development Llc | Methods and systems that convert document images to electronic documents using a trie data structure containing standard feature symbols to identify morphemes and words in the document images |
RU2556461C2 (en) * | 2013-11-18 | 2015-07-10 | Закрытое акционерное общество "ИнфоВотч" | Method for automated search of reference seals |
RU2641452C2 (en) * | 2013-12-18 | 2018-01-17 | ООО "Аби Девелопмент" | Incomplete standards |
JP6342298B2 (en) * | 2014-10-31 | 2018-06-13 | 株式会社東芝 | Character recognition device, image display device, image search device, character recognition method and program |
RU2605078C2 (en) * | 2015-01-27 | 2016-12-20 | Общество с ограниченной ответственностью "Аби Девелопмент" | Image segmentation for data verification |
RU2634192C1 (en) * | 2016-06-24 | 2017-10-24 | Общество с ограниченной ответственностью "Аби Девелопмент" | Data input from series of images applicable to template document |
JP2018018303A (en) * | 2016-07-28 | 2018-02-01 | 京セラドキュメントソリューションズ株式会社 | Information extraction device and image forming apparatus having the same |
US10417516B2 (en) | 2017-08-24 | 2019-09-17 | Vastec, Inc. | System and method for preprocessing images to improve OCR efficacy |
RU2657181C1 (en) * | 2017-09-01 | 2018-06-08 | Общество с ограниченной ответственностью "Аби Продакшн" | Method of improving quality of separate frame recognition |
RU2721186C1 (en) | 2019-07-22 | 2020-05-18 | Общество с ограниченной ответственностью "Аби Продакшн" | Optical character recognition of documents with non-planar regions |
CN110717489B (en) * | 2019-09-19 | 2023-09-15 | 平安科技(深圳)有限公司 | Method, device and storage medium for identifying text region of OSD (on Screen display) |
CN113239910B (en) * | 2021-07-12 | 2021-11-09 | 平安普惠企业管理有限公司 | Certificate identification method, device, equipment and storage medium |
CN113569864B (en) * | 2021-09-26 | 2022-01-25 | 广东电网有限责任公司中山供电局 | Method and system for identifying signature and seal |
US12008829B2 (en) | 2022-02-16 | 2024-06-11 | Vastec, Inc. | System and method for improved OCR efficacy through image segmentation |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5751850A (en) * | 1993-06-30 | 1998-05-12 | International Business Machines Corporation | Method for image segmentation and classification of image elements for documents processing |
US20020131642A1 (en) * | 2001-02-02 | 2002-09-19 | Lee Shih-Jong J. | Robust method for automatic reading of skewed, rotated or partially obscured characters |
US20060056696A1 (en) * | 2004-08-18 | 2006-03-16 | Fujitsu Limited | Degraded dictionary generation method and apparatus |
US20080069447A1 (en) * | 2006-09-14 | 2008-03-20 | Fujitsu Limited | Character recognition method, character recognition device, and computer product |
Family Cites Families (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP3272842B2 (en) * | 1992-12-17 | 2002-04-08 | ゼロックス・コーポレーション | Processor-based decision method |
ATE185211T1 (en) * | 1995-01-31 | 1999-10-15 | United Parcel Service Inc | METHOD AND DEVICE FOR SEPARATING THE FOREGROUND AND BACKGROUND IN IMAGES CONTAINING TEXT |
US5963666A (en) * | 1995-08-18 | 1999-10-05 | International Business Machines Corporation | Confusion matrix mediated word prediction |
US6507670B1 (en) * | 1998-03-05 | 2003-01-14 | Ncr Corporation | System and process for removing a background pattern from a binary image |
MC2491A1 (en) | 1999-06-21 | 1999-11-22 | Stringa Luigi | Automatic character recognition on a structured background by combining the background and character models |
US20030043172A1 (en) * | 2001-08-24 | 2003-03-06 | Huiping Li | Extraction of textual and graphic overlays from video |
RU2234126C2 (en) | 2002-09-09 | 2004-08-10 | Аби Софтвер Лтд. | Method for recognition of text with use of adjustable classifier |
US7650035B2 (en) * | 2006-09-11 | 2010-01-19 | Google Inc. | Optical character recognition based on shape clustering and multiple optical character recognition processes |
-
2008
- 2008-11-21 WO PCT/NO2008/000419 patent/WO2009070032A1/en active Application Filing
- 2008-11-21 RU RU2010125821/08A patent/RU2445699C1/en not_active IP Right Cessation
- 2008-11-21 US US12/744,974 patent/US8467614B2/en not_active Expired - Fee Related
- 2008-11-21 EP EP08855111A patent/EP2220590A1/en not_active Withdrawn
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5751850A (en) * | 1993-06-30 | 1998-05-12 | International Business Machines Corporation | Method for image segmentation and classification of image elements for documents processing |
US20020131642A1 (en) * | 2001-02-02 | 2002-09-19 | Lee Shih-Jong J. | Robust method for automatic reading of skewed, rotated or partially obscured characters |
US20060056696A1 (en) * | 2004-08-18 | 2006-03-16 | Fujitsu Limited | Degraded dictionary generation method and apparatus |
US20080069447A1 (en) * | 2006-09-14 | 2008-03-20 | Fujitsu Limited | Character recognition method, character recognition device, and computer product |
Non-Patent Citations (1)
Title |
---|
"Image Processing, 2001. International Conference on Proceedings, 2001", vol. 1, 7 October 2001, ISBN: 0-7803-6725-1, article QIGONG ZHENG ET AL.: "Morphological degradation models and their use in document image restoration, Image Processing, 2001.", pages: 193 - 196, XP010564829 * |
Also Published As
Publication number | Publication date |
---|---|
US8467614B2 (en) | 2013-06-18 |
RU2010125821A (en) | 2012-01-10 |
RU2445699C1 (en) | 2012-03-20 |
US20100303356A1 (en) | 2010-12-02 |
EP2220590A1 (en) | 2010-08-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US8467614B2 (en) | Method for processing optical character recognition (OCR) data, wherein the output comprises visually impaired character images | |
US7120318B2 (en) | Automatic document reading system for technical drawings | |
Shafait et al. | Table detection in heterogeneous documents | |
Guo et al. | Separating handwritten material from machine printed text using hidden markov models | |
US20110007366A1 (en) | System and method for classifying connected groups of foreground pixels in scanned document images according to the type of marking | |
Khurshid et al. | Word spotting in historical printed documents using shape and sequence comparisons | |
US20070065011A1 (en) | Method and system for collecting data from a plurality of machine readable documents | |
US10586133B2 (en) | System and method for processing character images and transforming font within a document | |
US8249364B2 (en) | Method for resolving contradicting output data from an optical character recognition (OCR) system, wherein the output data comprises more than one recognition alternative for an image of a character | |
Bhattacharya et al. | Understanding contents of filled-in Bangla form images | |
WO2007070010A1 (en) | Improvements in electronic document analysis | |
Khan et al. | Urdu optical character recognition technique using point feature matching; a generic approach | |
Tran et al. | A novel approach for text detection in images using structural features | |
Kumar et al. | Line based robust script identification for indianlanguages | |
Lin et al. | Multilingual corpus construction based on printed and handwritten character separation | |
JP4347675B2 (en) | Form OCR program, method and apparatus | |
JP4936250B2 (en) | Write extraction method, write extraction apparatus, and write extraction program | |
Höhn | Detecting arbitrarily oriented text labels in early maps | |
Kleber et al. | Document reconstruction by layout analysis of snippets | |
BJ et al. | Identification of Seal, Signature and Fingerprint from Malayalam Agreement Documents using Connected Component Analysis | |
Kronenberger et al. | Text and character recognition on metal-sheets | |
Arlandis et al. | Identification of very similar filled-in forms with a reject option | |
Sarkar et al. | A font invariant character segmentation technique for printed bangla word images | |
Sarkar et al. | Classifying foreground pixels in document images | |
Mandal et al. | Slant Estimation and Correction for Online Handwritten Bengali Words |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 08855111 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 982/MUMNP/2010 Country of ref document: IN |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
REEP | Request for entry into the european phase |
Ref document number: 2008855111 Country of ref document: EP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2008855111 Country of ref document: EP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2010125821 Country of ref document: RU |
|
WWE | Wipo information: entry into national phase |
Ref document number: 12744974 Country of ref document: US |