Abstract: A system and method for processing documents with automatic improvements to the processing. Documents are submitted to a processing system and data is extracted from the documents. The data may be extracted utilising OCR techniques. The data may be verified and interpreted utilising classifiers and predefined feature extraction rules which may improve their performance through an iterative learning cycle.

Abstract: A collection of labeled training cases is received, where each of the labeled training cases has at least one original feature and a label with respect to at least one class. Non-linear transformation of values of the original feature in the training cases is applied to produce transformed feature values that are more linearly related to the class than the original feature values. The non-linear transformation is based on computing probabilities of the training cases that are positive with respect to the at least one class. The transformed feature values are used to train a classifier.

Abstract: Architecture that performs image page index selection. A learning-based framework learns a statistical model based on the hyperlink (URL-uniform resource locator) previous click information obtained from the image search users. The learned model can combine the features of a newly discovered URL to predict the possibility of the newly-discovered URL being clicked in the future image search. In addition to existing web index selection features, image clicks are added as features, and the image clicks are aggregated over different URL segments, as well as the site modeling pattern trees to reduce the sparse problem of the image click information.

Abstract: A method of automatically generating a calibration file for a card handling device comprises automatically generating a calibration file stored in memory of a main control system for a card handling device. Automatically generating the calibration file comprises identifying at least one parameter associated with a rank area around a rank of the at least a portion of the card, identifying at least one parameter associated with a suit area around a suit of the at least a portion of the card, and storing the at least one parameter associated with the rank area and the at least one parameter associated with the suit area in the calibration file. Additionally, a method of automatically generating deck libraries for one or more decks of cards comprises automatically generate a plurality of master images for the cards of the first deck type using the parameters from the calibration file.

Abstract: A method and system for producing accented image data for an accented image is disclosed. The method includes decomposing each of a first and a second image into a gradient representation which comprises spectral and edge components. The first image comprises more spectral dimensions than the second image. The edge component from the first image is combined with the spectral component from the second image to form a combined gradient representation. Accented image data for the accented image is then generated from data including the combined gradient representation.

Abstract: An information processing apparatus, which creates a tree structure used by a recognition apparatus which recognizes specific information using the tree structure, including a memory unit which stores data including the information to be recognized and data not including the information so as to correspond to a label showing whether or not the data includes the information, a recognition device which recognizes the information and outputs a high score value when the data including the information is input, and a grouping unit which performs grouping of the recognition devices using a score distribution obtained when the data is input into the recognition devices.

Abstract: Methods, devices and systems are described for transcribing text from artifacts to electronic files. A computer system is provided, wherein the computer system comprises a computer-readable storage device. An image of the artifact is received wherein text is present on the artifact. A first portion of the text is analyzed. Characters representing the first portion of the text are identified at a first confidence level equal to or greater than a threshold confidence level. The characters representing the first portion of the text are stored. A second portion of the text appearing on the artifact is analyzed. A plurality of candidates to represent the second portion of the text are identified at a second confidence level below the threshold confidence level. Finally, the plurality of candidates to a user for selection are presented.

Abstract: Text in web pages or other text documents may be classified based on the images or other objects within the webpage. A system for identifying and classifying text related to an object may identify one or more web pages containing the image or similar images, determine topics from the text of the document, and develop a set of training phrases for a classifier. The classifier may be trained and then used to analyze the text in the documents. The training set may include both positive examples and negative examples of text taken from the set of documents. A positive example may include captions or other elements directly associated with the object, while negative examples may include text taken from the documents, but from a large distance from the object. In some cases, the system may iterate on the classification process to refine the results.

Abstract: Systems and methods for improving visual object recognition by analyzing query images are disclosed. In one example, a visual object recognition module may determine query images matching objects of a training corpus utilized by the module. Matched query images may be added to the training corpus as training images of a matched object to expand the recognition of the object by the module. In another example, relevant candidate image corpora from a pool of image data may be automatically selected by matching the candidate image corpora against user query images. Selected image corpora may be added to a training corpus to improve recognition coverage. In yet another example, objects unknown to a visual object recognition module may be discovered by clustering query images. Clusters of similar query images may be annotated and added into a training corpus to improve recognition coverage.

Abstract: Techniques and methods are disclosed herein for combining and weighting of values from and associated with classifiers. Classifiers are used to recognize characters as part of an optical character recognition (OCR) system. Various methods of normalization facilitate combining of results of classifiers. For example, weight values may be entered into a weight table having two columns, one that includes weights from comparing patterns with images of correct characters, the other column includes weights from comparing patterns with images of incorrect characters.

Abstract: A character recognition system receives an unknown character and recognizes the character based on a pre-trained recognition model. Prior to recognizing the character, the character recognition system may pre-process the character to rotate the character to a normalized orientation. By rotating the character to a normalized orientation in both training and recognition stages, the character recognition system releases the pre-trained recognition model from considering character prototypes in different orientations and thereby speeds up recognition of the unknown character. In one example, the character recognition system rotates the character to the normalized orientation by aligning a line between a sum of coordinates of starting points and a sum of coordinates of ending points of each stroke of the character with a normalized direction.

Abstract: An information recognition system includes: a display section displaying an image on a display surface at a predetermined display resolution; an image combining section combining a character entry guide with the image, the character entry guide assisting handwritten input to the display surface; an information detecting section detecting handwritten input information at a detection resolution which is higher than the display resolution, the handwritten input information input to the display surface according to the character entry guide; and a character recognizing section performing character recognition based on the information detected at the detection resolution.

Abstract: A method for automatically recognizing Arabic text includes digitizing a line of Arabic characters to form a two-dimensional array of pixels each associated with a pixel value, wherein the pixel value is expressed in a binary number, dividing the line of the Arabic characters into a plurality of line images, defining a plurality of cells in one of the plurality of line images, wherein each of the plurality of cells comprises a group of adjacent pixels, serializing pixel values of pixels in each of the plurality of cells in one of the plurality of line images to form a binary cell number, forming a text feature vector according to binary cell numbers obtained from the plurality of cells in one of the plurality of line images, and feeding the text feature vector into a Hidden Markov Model to recognize the line of Arabic characters.

Abstract: An apparatus for pattern processing exhibits a discretizing device for discretizing an input pattern, a device for generating a number n of discrete variants of the quantized input pattern in accordance with established rules, a number n of input stages (50) for generating, for each input-pattern variant, an assigned output symbol from a set of symbols, and a selection unit (60) for selecting a symbol by way of selected symbol relating to the input pattern from the n generated output symbols in accordance with an established selection rule. The apparatus according to the invention and the corresponding process according to the invention enable a faster, more precise and more flexible recognition of patterns, in which connection it may be a question of spatial image patterns, temporally variable signal patterns and other input patterns.

Abstract: An image display apparatus is disclosed. The image display apparatus includes a detection section, an image forming section, and a display process section. The detection section detects a user's watching state. The image forming section that forms a display image which is displayed on a screen based on a plurality of images and changes the display image based on a detected result of the detection section. The display process section which performs a process of displaying the display image formed by the image forming section.

Abstract: One or more techniques and/or systems are disclosed for compensating for affine distortions in handwriting recognition. Orientation estimation is performed on a handwriting sample to generate a set of likely characters for the sample. An estimated affine transform is determined for the sample by applying hidden Markov model (HMM) based minimax testing to the sample using the set of likely characters. The estimated affine transform is applied to the sample to compensate for the affine distortions of the sample, yielding an affine distortion compensated sample.

Abstract: Methods and systems of the present embodiment provide segmenting of connected components of markings found in document images. Segmenting includes detecting aligned text. From this detected material an aligned text mask is generated and used in processing of the images. The processing includes breaking connected components in the document images into smaller pieces or fragments by detecting and segregating the connected components and fragments thereof likely to belong to aligned text.

Abstract: Embodiments of the invention describe methods and apparatus for performing context-sensitive OCR. A device obtains an image using a camera coupled to the device. The device identifies a portion of the image comprising a graphical object. The device infers a context associated with the image and selects a group of graphical objects based on the context associated with the image. Improved OCR results are generated using the group of graphical objects. Input from various sensors including microphone, GPS, and camera, along with user inputs including voice, touch, and user usage patterns may be used in inferring the user context and selecting dictionaries that are most relevant to the inferred contexts.

Abstract: A system, method and computer program product are provided for detecting unwanted data. In use, data is rendered, after which it may be determined whether the rendered data is unwanted, utilizing either a neural network or optical character recognition.

Abstract: Systems and methods are provided for evaluating and sorting seeds based on characteristics of the seeds. One system includes an imaging and analysis subsystem that collects image data from the seeds and analyzes the collected image data for characteristics of the seeds. This subsystem can include an imaging theater having mirrors that reflect image data from the seeds to an imaging device for collection. The system can also include an off-loading and sorting subsystem configured to sort the seeds based on their characteristics. And, one method includes illuminating the seeds and collecting image data from the seeds for determining their characteristics. The image data can be collected from at least three portions of the seeds at each of a plurality of sequentially changing spectral wavelengths. In addition (or alternatively), the image data can be collected from top and bottom portions of the seeds using a single imaging device.

Abstract: Techniques and methods are disclosed herein for combining and weighting of values from and associated with classifiers. Classifiers are used to recognize characters as part of an optical character recognition (OCR) system. Various methods of normalization facilitate combining of results of classifiers. For example, weight values may be entered into a weight table having two columns, one that includes weights from comparing patterns with images of correct characters, the other column includes weights from comparing patterns with images of incorrect characters.

Abstract: Described is a technology in which a low resolution image is processed into a high-resolution image, including by a two interpolation passes. In the first pass, missing in-block pixels, which are the pixels within a block formed by four neighboring original pixels, are given values by gradient diffusion based upon interpolation of the surrounding original pixels. In the second interpolation pass, missing on-block pixels, which are the pixels on a block edge formed by two adjacent original pixels, are given values by gradient diffusion based upon interpolation of the values of those adjacent original pixels and the previously interpolated values of their adjacent in-block pixels. Also described is a difference projection process that varies the values of the interpolated pixels according to a computed difference projection.

Abstract: Systems and methods for implementing a multi-label image recognition framework for classifying digital images are provided. The provided multi-label image recognition framework utilizes an iterative, multiple analysis path approach to model training and image classification tasks. A first iteration of the multi-label image recognition framework generates confidence maps for each label, which are shared by the multiple analysis paths to update the confidence maps in subsequent iterations. The provided multi-label image recognition framework permits model training and image classification tasks to be performed more accurately than conventional single-label image recognition frameworks.

Abstract: One or more techniques and/or systems are disclosed for constructing a compact handwriting character classifier. A precision constrained Gaussian model (PCGM) based handwriting classifier is trained by estimating parameters for the PCGM under minimum classification error (MCE) criterion, such as by using a computer-based processor. The estimated parameters of the trained PCGM classifier are compressed using split vector quantization (VQ) (e.g., and in some embodiments, scalar quantization) to compact the handwriting recognizer in computer-based memory.

Abstract: A method for automatically recognizing Arabic text includes digitizing a line of Arabic characters to form a two-dimensional array of pixels each associated with a pixel value, wherein the pixel value is expressed in a binary number, dividing the line of the Arabic characters into a plurality of line images, defining a plurality of cells in one of the plurality of line images, wherein each of the plurality of cells comprises a group of adjacent pixels, serializing pixel values of pixels in each of the plurality of cells in one of the plurality of line images to form a binary cell number, forming a text feature vector according to binary cell numbers obtained from the plurality of cells in one of the plurality of line images, and feeding the text feature vector into a Hidden Markov Model to recognize the line of Arabic characters.

Abstract: Product images are used in conjunction with textual descriptions to improve classifications of product offerings. By combining cues from both text and image descriptions associated with products, implementations enhance both the precision and recall of product description classifications within the context of web-based commerce search. Several implementations are directed to improving those areas where text-only approaches are most unreliable. For example, several implementations use image signals to complement text classifiers and improve overall product classification in situations where brief textual product descriptions use vocabulary that overlaps with multiple diverse categories. Other implementations are directed to using text and images “training sets” to improve automated classifiers including text-only classifiers.

Abstract: An image processing device is provided which generates an easily reusable electronic document from an input image in which different page sizes are mixed. The image processing device generates a plurality of pieces of display information from a plurality of document images, and, depending on the size and the direction of each of the images, converts the pieces of display information into electronic documents. That is, the plurality of pieces of display information are divided into a plurality of groups, depending on the size and the direction of each of the images, and the display information included in each of the groups is converted into a separate electronic document. Further, sequence information based on the input order of the plurality of document images is stored on an electronic document.

Abstract: Various technologies and techniques are disclosed that improve handwriting recognition operations. Handwritten input is received in training mode and run through several base recognizers to generate several alternate lists. The alternate lists are unioned together into a combined alternate list. If the correct result is in the combined list, each correct/incorrect alternate pair is used to generate training patterns. The weights associated with the alternate pairs are stored. At runtime, the combined alternate list is generated just as training time. The trained comparator-net can be used to compare any two alternates in the combined list. A template matching base recognizer is used with one or more neural network base recognizers to improve recognition operations. The system provides comparator-net and reorder-net processes trained on print and cursive data, and ones that have been trained on cursive-only data. The respective comparator-net and reorder-net processes are used accordingly.

Abstract: A difference emphasizing apparatus aligns a first three-dimensional model and a second three-dimensional model in orientation and position in accordance with a predetermined rule, and gets data of respective apices of the first three-dimensional model and the second three-dimensional model. Based on the gotten data, the apparatus finds a corresponding point on the first three-dimensional model, which corresponds to the apex of the second three-dimensional model in a direction of a particular axis. When the corresponding point is detected, the apparatus calculates a difference between the first three-dimensional model and the second three-dimensional model in the direction of the particular axis based on the corresponding point and the apex of the second three-dimensional model. The apparatus enlarges the difference in the direction of the particular axis, and calculates a position of the apex of the second three-dimensional model after the enlargement.

Abstract: An image processing method is used to detect a letter by using a classifier generated through statistical learning of handling a sample image of a fixed size as supervised data, and includes the following steps. A conversion step acquires a converted image by geometrically converting a target image containing a letter to be detected such that the target image has a predetermined ratio defining an aspect ratio. A search step searches the converted image for one or more letter candidates each including a region of a possible letter by using the classifier. An integration step applies clustering to the letter candidates, integrating the letter candidates, and eliminates the letter candidate having low reliability A circumscribing step cuts a letter out of the letter candidate that has been integrated and has not been eliminated, and generates a rectangle circumscribing the letter.

Abstract: In a method for acquiring data from a machine-readable document for assignment to fields of a database, individual data are extracted substantially automatically from the document and entered into the corresponding database fields. If data cannot be extracted from the document with a desired degree of reliability for one or more particular database fields, then the steps are executed of displaying the document onto the display screen, displaying on the display screen the at least one or more database fields for which the data cannot be extracted with the desired degree of reliability, and executing a proposal routine with which string sections in the vicinity of a pointer movable by a user on the display screen are selected, marked, and proposed for extraction.

Abstract: Various technologies and techniques are disclosed that improve cursive handwriting recognition. Cursive handwriting input is received from a user. The system performs a hierarchical prototype search as part of a recognition operation. A same space search is performed against a mixed database that has both print and cursive samples. A same space search is also performed against a cursive database that has only cursive samples. The results of these two same space searches are merged into a combined alternate list. The combined alternate list is then used as a constraint for the dynamic time warp searches that are performed against the mixed and cursive databases, respectively. The results of the dynamic time warp searches are also merged into a final combined alternate list, and the combined alternate list is used to make a recognition decision regarding the user's handwritten input.

Abstract: One or more techniques and/or systems are disclosed for mitigating machine solvable human interactive proofs (HIPs). A classifier is trained over a set of one or more training HIPs that have known characteristics for OCR solvability and HIP solving pattern from actual use. A HIP classification is determined for a HIP (such as from a HIP library used by a HIP generator) using the trained classifier. If the HIP is classified by the trained classifier as a merely human solvable classification, such that it may not be solved by a machine, the HIP can be identified for use in the HIP generation system. Otherwise, the HIP can be altered to (attempt to) be merely human solvable.

Abstract: A method for automatically recognizing Arabic text includes digitizing a line of Arabic characters to form a two-dimensional array of pixels each associated with a pixel value, wherein the pixel value is expressed in a binary number, dividing the line of the Arabic characters into a plurality of line images, defining a plurality of cells in one of the plurality of line images, wherein each of the plurality of cells comprises a group of adjacent pixels, serializing pixel values of pixels in each of the plurality of cells in one of the plurality of line images to form a binary cell number, forming a text feature vector according to binary cell numbers obtained from the plurality of cells in one of the plurality of line images, and feeding the text feature vector into a Hidden Markov Model to recognize the line of Arabic characters.

Abstract: Input personal handwriting of a character stored in a system character database into an input device. Divide the personal handwriting of the character into a group of personalized roots. Store the group of personalized roots in a personalized-root database. Form a plurality of personalized characters according to a plurality of personalized roots stored in the personalized-root database. Store the plurality of personalized characters in a personalized-character database.

Abstract: One or more techniques and/or systems are disclosed for compensating for affine distortions in handwriting recognition. Orientation estimation is performed on a handwriting sample to generate a set of likely characters for the sample. An estimated affine transform is determined for the sample by applying hidden Markov model (HMM) based minimax testing to the sample using the set of likely characters. The estimated affine transform is applied to the sample to compensate for the affine distortions of the sample, yielding an affine distortion compensated sample.

Abstract: One or more techniques and/or systems are disclosed for constructing a compact handwriting character classifier. A precision constrained Gaussian model (PCGM) based handwriting classifier is trained by estimating parameters for the PCGM under minimum classification error (MCE) criterion, such as by using a computer-based processor. The estimated parameters of the trained PCGM classifier are compressed using split vector quantization (VQ) (e.g., and in some embodiments, scalar quantization) to compact the handwriting recognizer in computer-based memory.

Abstract: A data processing apparatus includes first and second unsupervised learning process units and a supervised learning process unit. The first unsupervised learning process unit classifies data of a first data group according to unsupervised learning, to perform dimension reduction for the first data group and to obtain first classified data. The second unsupervised learning process unit classifies data of a second data group according to the unsupervised learning, to perform dimension reduction for the second data group and to obtain a second classified data group. The supervised learning process unit performs supervised learning using, as a teacher, the first classified data group obtained by the first unsupervised learning process unit and the second classified data group obtained by the second unsupervised learning process unit to determine a mapping relation between the first classified data group and the second classified data group.

Abstract: An exemplary method for handwritten character generation includes receiving one or more characters and, for the one or more received characters, generating handwritten characters using Hidden Markov Models trained for generating handwritten characters. In such a method the trained Hidden Markov Models can be adapted using a technique such as a maximum a posterior technique, a maximum likelihood linear regression technique or an Eigen-space technique.

Abstract: A system for recording handwritten notes includes a feature information obtaining section that obtains feature information of a user who is holding a handwriting tool, a handwritten notes obtaining section that obtains notes handwritten with the handwriting tool, and a recording section that records the feature information of the user who is holding the handwriting tool and the notes handwritten with the handwriting tool, the handwritten notes being directly or indirectly associated with the feature information.

Abstract: Methods, devices and systems are described for transcribing text from artifacts to electronic files. A computer system is provided, wherein the computer system comprises a computer-readable storage device. An image of the artifact is received wherein text is present on the artifact. A first portion of the text is analyzed. Characters representing the first portion of the text are identified at a first confidence level equal to or greater than a threshold confidence level. The characters representing the first portion of the text are stored. A second portion of the text appearing on the artifact is analyzed. A plurality of candidates to represent the second portion of the text are identified at a second confidence level below the threshold confidence level. Finally, the plurality of candidates to a user for selection are presented.

Abstract: A forward pass through a sequence of strokes representing a handwritten equation is performed from the first stroke to the last stroke in the sequence. At each stroke, a path score is determined for a plurality of symbol-relation pairs that each represents a symbol and its spatial relation to a predecessor symbol. A symbol graph having nodes and links is constructed by backtracking through the strokes from the last stroke to the first stroke and assigning scores to the links based on the path scores for the symbol-relation pairs. The symbol graph is used to recognize a mathematical expression based in part on the scores for the links and the mathematical expression is stored.

Abstract: A character input method using a touch screen, in which one or more areas requiring user input is defined in the touch screen, pre-recognized information is defined for each of the defined areas, character information is received by a user in one or more user desired areas among the defined areas, the character information is recognized using a character recognizer, and the recognized character information is updated in the user desired areas.

Abstract: An electronic endoscope apparatus includes an imaging element and a signal processing unit. The imaging element obtains an image of an observation object, and outputs an image signal of the observation object. The signal processing unit alternately repeats obtainment of a partial image signal using a part of a light receiving area of the imaging element and obtainment of a partial image signal using the remaining part of the light receiving area. The signal processing unit also obtains a whole image signal corresponding to an image of the observation object using a partial image signal obtained in the n-th (n is a natural number) obtainment and a partial image signal obtained in the (n+1)th obtainment. Further, a partial component of the n-th partial image signal is extracted by an extraction unit, and the extracted partial component is added to the (n+1)th partial image signal.

Abstract: A method and system for automatically training a document imaging classification and extraction system that switches between a manual mode and an automatic mode based on constant monitoring. A specialized sub-system monitors and records a user interaction with the classification system during the initial manual mode and, in parallel, develops and tests a user configuration with respect to an automated processing engine. The system is capable of being shifted to the automatic mode if a desired acceptability threshold is attained and the document can then be processed automatically. Furthermore, a user can interact with the classification system if the automatic mode fails. Information concerning exception handling can be entered into a training database for continual refinement of the classification and extraction system.

Abstract: A system and method are disclosed for learning a random multinomial logit (RML) classifier and applying the RML classifier for scene segmentation. The system includes an image textonization module, a feature selection module and a RML classifier. The image textonization module is configured to receive an image training set with the objects of the images being pre-labeled. The image textonization module is further configured to generate corresponding texton images from the image training set. The feature selection module is configured to randomly select one or more texture-layout features from the texton images. The RML classifier comprises multiple multinomial logistic regression models. The RML classifier is configured to learn each multinomial logistic regression model using the selected texture-layout features. The RML classifier is further configured to apply the learned regression models to an input image for scene segmentation.

Abstract: Described is a technology by which handwriting recognition is performed using a precision constrained Gaussian model (PCGM) that requires far less memory than other models such as MQDF. Offline training, such as via maximum likelihood and/or minimum classification error techniques, provides classification data. The classification data includes basis matrices that are shared by classes, along with weighting coefficients and a mean vector corresponding to each class. The base matrices and weights are obtained by expanding a precision matrix for each class. In online recognition, received handwritten input (e.g., an East Asian character) is classified into a class, based upon the per-class mean vector and weighting coefficients, and the basis matrices, by a PCGM recognizer that outputs similarity scores for candidates and a decision rule that selects the most likely class.

Abstract: A source keyword may be received multiple times and each time, in response, a machine-learning algorithm may be used to identify and rank respective matching-keywords that have been determined to match the source keyword. A portion or unit of content may be generated based on one of the ranked matching-keywords. The content is transmitted via a network to a client device and a user's impression of the content is recorded. The machine-learning algorithm may continue to rank matching-keywords for arbitrary source keywords while the recorded impressions and corresponding matched-keywords, respectively, are used to train the machine-learning algorithm. The training alters how the machine-learning algorithm ranks matching-keywords determined to match the source keyword.

Abstract: The present invention discloses methods and systems for discovering printers and shares on a computer network. Each domain on the network is identified, and each computer in the domain is identified. In addition, each printer connected to the computer and each share on the computer is identified. Shortcuts to the identified printers and shares are created on at least one computer on the network. Moreover, drivers are preferably installed on the computer for each printer for which a shortcut was created. In the event that the total number of resources (i.e., shares and/or printers) exceeds a threshold, then the process terminates. Otherwise, the present invention continues until all printers and shares on the network are identified, and the appropriate shortcuts are created. Thus, the present invention provides methods and systems for discovering resources on a network.

Abstract: An image-recognition method and a system using the method is disclosed, which proceeds comparison through the visual lingual characteristics according to the logic of lingual vocabulary of an image to be recognized to reduce the number of the objects to be compared, and to select at least one object. After that, similarity comparison between a graphic characteristic of the image to be recognized and at least one graphic sample corresponding to the object selected is proceeded. And then, at least one graphic sample is selected to achieve open frame image recognition.