Abstract: Methods, and systems, including computer programs encoded on computer-readable mediums, including a method for determining codomain relationship measures between domains pairs. The method includes accessing domain data specifying first and second domains; for each first domain: determining a total number of image links on web pages on the first domain, each image link being a link to an image on a second domain; identifying second domains hosting images to which image links on web pages on the first domain link; for each second domain: determining a number of image links from web pages on the first domain linking to images on the second domain; determining a codomain relationship measure for the first and second domain based on the number of image links from web pages on the first domain linking to images on the second domain and the total number of image links on web pages on the first domain.

Abstract: Methods and system employing the same for optimizing an objective function are provided. The objective function assesses the quality of a candidate solution. One or more variables of an objective function are selected as pivot variables. Each of the variables include one or more candidate values. An upper bound function is generated from the objective function, where the pivot variables are held fixed. For each combination of the candidate values, one or more candidate solutions are searched using the upper bound function. One or more optimal solutions are selected from among the solutions to the searches.

Abstract: Methods and system employing the same for finding repeated structure for data extraction from document images are provided. A reference record and one or more reference fields thereof are identified from a document image. One or more candidate fields are generated for each of the reference fields. One or more best candidate records from the candidate fields are selected using a probabilistic model and an optimal record set is determined from the best candidate records.

Abstract: A user interface and method is embodied on a computer readable medium and executable on a computer. The user interface is a labeler which labels only foreground pixels of an image stored in a computing environment. The labeler operates in a Region mode/state and Brush mode/state, and includes a Tentative mode that permits an assigned label to be changed after pixels have been selected. Groups of pixels may be selected for labeling at once by a point-and-click command, and a pixel may belong to one or more groups of pixels which are stored in memory as image layers. The groups are formed dynamically by user selection actions, and/or through automatic recognition algorithms. Pixels already labeled with certain labels may be locked to not be altered by additional labeling operations. Unassigned pixels may be highlighted to increase the ease at which they are identified in an image. Comparisons between labeled images are undertaken to indicate differences between different groundtruth labeling.

Abstract: Methods and systems for classifying markings on images in a document are undertaken according to marking types. The document containing the images is supplied to a segmenter which breaks the images into fragments of foreground pixel structures that are identified as being likely to be of the same marking type by finding connected components, extracting near-horizontal or -vertical rule lines and subdividing some connected components to obtain the fragments. The fragments are then supplied to a classifier, where the classifier provides a category score for each fragment, wherein the classifier is trained from the groundtruth images whose pixels are labeled according to known marking types. Thereafter, a same label is assigned to all pixels in a particular fragment, when the fragment is classified by the classifier.

Abstract: The present application is a method and system of interpreting an image by finding a configuration of multiple variables which optimizes an objective function with a factorizable upper bound, by applying an iterative algorithm that relies on efficient dynamic ordering of candidate configurations, in a priority queue, in a descending order of an upper bound score. As an example, consider a constellation model for an object. It specifies the appearance models for individual parts of objects, as well as spatial relations among these parts. These are combined into a single function whose value represents the likeness of the object in an image. To find the configuration in which the object is present in the image, we maximize this function over all candidate configurations. The purpose of the iterative algorithm mentioned above is to find such optimal configurations efficiently.

Abstract: Methods of generating image anchor templates from low variance regions of document images of a first class are provided. The methods select a document image from the document images of the first class and align the other document images of the first class to the selected document image. Low variance regions are then determined by comparing the aligned document images and the selected document image and used to generate image anchor templates.

Abstract: Methods, and corresponding systems, of generating one or more image anchor templates for extracting data from a data field of a first class of documents are provided. The methods include generating one or more candidate image anchor templates from at least one of one or more exemplars of the first class; determining a quality score for each of the one or more candidate image anchor templates using a computer processor and known locations of the data field within the one or more exemplars of the first class; ranking the one or more candidate image anchor templates according to quality score; and selecting one or more of the most highly ranked image anchor templates.

Abstract: Methods, and corresponding systems, of generating one or more image anchor templates for discriminating between documents of a first class and documents of other classes are provided. The methods include generating one or more candidate image anchor templates; determining, using a computer processor, a quality score for each of the one or more candidate image anchor templates; ranking the one or more candidate image anchor templates according to the quality scores of the one or more candidate image anchor templates; and selecting one or more of the most highly ranked image anchor templates.

Abstract: Methods and system employing the same for finding repeated structure for data extraction from document images are provided. A reference record and one or more reference fields thereof are identified from a document image. One or more candidate fields are generated for each of the reference fields. One or more best candidate records from the candidate fields are selected using a probabilistic model and an optimal record set is determined from the best candidate records.

Abstract: Methods and system employing the same for optimizing an objective function are provided. The objective function assesses the quality of a candidate solution. One or more variables of an objective function are selected as pivot variables. Each of the variables include one or more candidate values. An upper bound function is generated from the objective function, where the pivot variables are held fixed. For each combination of the candidate values, one or more candidate solutions are searched using the upper bound function. One or more optimal solutions are selected from among the solutions to the searches.

Abstract: Methods of generating image anchor templates from low variance regions of document images of a first class are provided. The methods select a document image from the document images of the first class and align the other document images of the first class to the selected document image. Low variance regions are then determined by comparing the aligned document images and the selected document image and used to generate image anchor templates.

Abstract: The present application is a method and system of interpreting an image by finding a configuration of multiple variables which optimizes an objective function with a factorizable upper bound, by applying an iterative algorithm that relies on efficient dynamic ordering of candidate configurations, in a priority queue, in a descending order of an upper bound score. As an example, consider a constellation model for an object. It specifies the appearance models for individual parts of objects, as well as spatial relations among these parts. These are combined into a single function whose value represents the likeness of the object in an image. To find the configuration in which the object is present in the image, we maximize this function over all candidate configurations. The purpose of the iterative algorithm mentioned above is to find such optimal configurations efficiently.

Abstract: Methods, and corresponding systems, of generating one or more image anchor templates for discriminating between documents of a first class and documents of other classes are provided. The methods include generating one or more candidate image anchor templates; determining, using a computer processor, a quality score for each of the one or more candidate image anchor templates; ranking the one or more candidate image anchor templates according to the quality scores of the one or more candidate image anchor templates; and selecting one or more of the most highly ranked image anchor templates.

Abstract: Methods, and corresponding systems, of generating one or more image anchor templates for extracting data from a data field of a first class of documents are provided. The methods include generating one or more candidate image anchor templates from at least one of one or more exemplars of the first class; determining a quality score for each of the one or more candidate image anchor templates using a computer processor and known locations of the data field within the one or more exemplars of the first class; ranking the one or more candidate image anchor templates according to quality score; and selecting one or more of the most highly ranked image anchor templates.

Abstract: A system and method for image performing classification through generative models of features occurring in an image. Category-conditional probability distributions of features occurring in a plurality of training images are maintained. Each distribution is assigned to a category. The features occurring in an unclassified image are identified. Category-conditional likelihoods for the identified features are determined using the category-conditional probability distributions for each category. The unclassified image is assigned to one such category based upon the category-conditional likelihoods.

Abstract: Methods and systems for classifying markings on images in a document are undertaken according to marking types. The document containing the images is supplied to a segmenter which breaks the images into fragments of foreground pixel structures that are identified as being likely to be of the same marking type by finding connected components, extracting near-horizontal or -vertical rule lines and subdividing some connected components to obtain the fragments. The fragments are then supplied to a classifier, where the classifier provides a category score for each fragment, wherein the classifier is trained from the groundtruth images whose pixels are labeled according to known marking types. Thereafter, a same label is assigned to all pixels in a particular fragment, when the fragment is classified by the classifier.

Abstract: A user interface and method is embodied on a computer readable medium and executable on a computer. The user interface is a labeler which labels only foreground pixels of an image stored in a computing environment. The labeler operates in a Region mode/state and Brush mode/state, and includes a Tentative mode that permits an assigned label to be changed after pixels have been selected. Groups of pixels may be selected for labeling at once by a point-and-click command, and a pixel may belong to one or more groups of pixels which are stored in memory as image layers. The groups are formed dynamically by user selection actions, and/or through automatic recognition algorithms. Pixels already labeled with certain labels may be locked to not be altered by additional labeling operations. Unassigned pixels may be highlighted to increase the ease at which they are identified in an image. Comparisons between labeled images are undertaken to indicate differences between different groundtruth labeling.

Abstract: A system and method for image performing classification through generative models of features occurring in an image. Category-conditional probability distributions of features occurring in a plurality of training images are maintained. Each distribution is assigned to a category. The features occurring in an unclassified image are identified. Category-conditional likelihoods for the identified features are determined using the category-conditional probability distributions for each category. The unclassified image is assigned to one such category based upon the category-conditional likelihoods.

Abstract: Techniques are provided to classify patterns in isogenous pattern sources. Techniques are provided to determine a computationally inexpensive upperbound on the true score or joint probability of the field label and field features over all field labels. Candidate field labels associated with promising upperbound scores are dynamically queued. True scores are computed for a subset of the candidates fields resulting in reduced computations to determine a field label. Techniques are also provided to determine optimal variables for any system with shared constraints.