Abstract: A programming interface of document search system enables a user to dynamically specifying features of documents recorded in a corpus of documents. The programming interface provides category and format flexibility for defining different genre of documents. The document search system initially segments document images into one or more layout objects. Each layout object identifies a structural element in a document such as text blocks, graphics, or halftones. Subsequently, the document search system computes a set of attributes for each of the identified layout objects. The set of attributes are used to describe the layout structure of a page image of a document in terms of the spatial relations that layout objects have to frames of reference that are defined by other layout objects. Using the set of attributes a user defines features of a document with the programming interface.

Abstract: The spatial separations between clusters of objects on the electronic work surface are analyzed to automatically generate borders, which may be used in at least two manners. The borders may become explicit editable borders within an output image. Alternatively, the borders created may be implicit ephemeral borders. According to an aspect, a method uses the sufficient stability grouping technique to determine grouping of elements in the input image array into the text lines in the output text lines array. According to another aspect, a method uses the sufficient stability grouping technique to determine the amount of vertical expansion to apply to each text line so as to create text regions in the output array which correspond to the text lines in the input text lines array. According to yet another aspect, a method takes the text regions array as input and extracts enclosure borders from the text regions.

Abstract: A programming interface of document search system enables a user to dynamically specifying features of documents recorded in a corpus of documents. The programming interface provides category and format flexibility for defining different genre of documents. The document search system initially segments document images into one or more layout objects. Each layout object identifies a structural element in a document such as text blocks, graphics, or halftones. Subsequently, the document search system computes a set of attributes for each of the identified layout objects. The set of attributes are used to describe the layout structure of a page image of a document in terms of the spatial relations that layout objects have to frames of reference that are defined by other layout objects. Using the set of attributes a user defines features of a document with the programming interface.

Abstract: The spatial separations between clusters of objects on the electronic work surface are analyzed to automatically generate borders, which may be used in at least two manners. The borders may become explicit editable borders within an output image. Alternatively, the borders created may be implicit ephemeral borders. According to an aspect, a method uses the sufficient stability grouping technique to determine grouping of elements in the input image array into the text lines in the output text lines array. According to another aspect, a method uses the sufficient stability grouping technique to determine the amount of vertical expansion to apply to each text line so as to create text regions in the output array which correspond to the text lines in the input text lines array. According to yet another aspect, a method takes the text regions array as input and extracts enclosure borders from the text regions.

Abstract: A method and apparatus for compressing a corpus of document images into a collective tokenized representation. Initially, documents in the corpus are individually compressed into a document tokenized format. A document image in the document tokenized format is represented using a symbol table and a table of positions. Each symbol in the symbol table is a shape in the original document image. The positions in the table of positions indicates where the symbols in the symbol table are placed to form the document image. Subsequently, the individual symbol tables of each document in the corpus are assembled to form clusters of similar shapes. These clusters are then analyzed to identify the degree of interrelationship between the symbols in the individual symbol tables. Individual document symbol tables with a large number of recurring symbols are grouped together. For each of the groups of symbol tables, a collective symbol table is computed.

Abstract: The present invention is a method for analyzing image data, and more particularly for analyzing of image data representing images containing text to partition the image into running and non-running text regions and to further classify the non-running text regions therein. The present invention utilizes characteristics of running text regions to identify such regions and to subsequently group all non-running text regions into related groups prior to the classification of the non-running text regions. Classification of the non-running text regions is accomplished by analyzing whether the non-running text regions exhibit pronounced horizontal and/or vertical alignment of the text blocks therein. Once the analysis is complete, alignment information is used to determine the number of "rows" and "columns" so as to classify the non-running text region as text, a horizontal sequence, a vertical sequence, or a table.

Abstract: A document search system provides a user with a programming interface for dynamically specifying features of documents recorded in a corpus of documents. The programming interface operates at a high-level that is suitable for interactive user specification of layout components and structures of documents. In operation, a bitmap image of a document is analyzed by the document search system to identify layout objects such as text blocks or graphics. Subsequently, the document search system computes a set of attributes for each of the identified layout objects. The set of attributes which are identified are used to describe the layout structure of a page image of a document in terms of the spatial relations that layout objects have to frames of reference that are defined by other layout objects. After computing attributes for each layout object, a user can operate the programming interface to define unique document features.

Abstract: The present invention is a method and apparatus for analyzing image data, and more particularly for analyzing image data representing images containing text to partition the image into running and non-running text regions therein. The present invention utilizes characteristics of running text regions to identify such regions and to subsequently group all non-running text regions into related groups.

Abstract: Input image data define an input image set that shows a graphical feature and editing marks indicating an editing operation to be performed on the graphical feature. The input image data are used to obtain operation category data indicating whether the editing operation would translate the graphical feature so that it is centered at a different position within the input image set. The operation category data are used to obtain output image data defining an output image that includes an edited version of the input image set. The output image shows the graphical feature centered at a different position only if the operation category data so indicate. The input image set can include an original image showing the graphical feature and an overlay image showing the editing marks. The editing marks can form a node-link structure with the graphical feature.

Abstract: Input image data define an input image that shows a perimeter relationship representation, such as a Venn diagram or statechart. The representation includes a perimeters feature that satisfies a constraint on perimeters. The constraint can include a perimeter size criterion that distinguishes perimeters from labels. Or the constraint can include an enclosing perimeter criterion requiring a connected component within each perimeter and a perimeter label criterion requiring a label for each perimeter. The constraint can also include an empty perimeter criterion to distinguish empty perimeters from labels. The input image data are used to obtain perimeters data indicating parts of the input image that satisfy the constraint. The perimeters data are used to obtain relationship data indicating a relationship between distinctions represented by the perimeters.

Abstract: Image data define an image that shows an input graphical representation whose configuration represents information. The image can show a sketch, for example. The image data are used to obtain category data indicating a category of graphical representations. For example, the category data could indicate any of X-Y graph, directed graph, undirected graph, bar graph, pie chart, line graph, scatter plot, circuit diagram, flow chart, Venn diagram, state-transition diagram, tree, table, matrix, or array. The category data could also indicate a specific category within a generic category such as bar graph or pie chart. The category data can be obtained by determining whether the graphical representation satisfies a constraint on each category.

Abstract: To group items in an array, gap data are obtained indicating gaps between items. The gap data are used to obtain threshold data, which are then used to obtain grouping data. The gaps could, for example, be distances between items in a two-dimensional array or differences between values at which items occur in a one-dimensional array. The threshold data indicate a threshold. The threshold would produce a number of groups of the items that is stable across a range of thresholds, and the range of thresholds meets a criterion for largeness of a range. The criterion can require, for example, that the range be larger than the stable range of thresholds of any other number in a set of numbers of groups. The threshold can be obtained iteratively by applying a candidate threshold for each iteration. The candidate thresholds can be incremented, and the iterations can be counted to find a number of groups meeting the criterion.

Abstract: Input image data define an input image set that shows a node-link structure, such as a directed graph, an undirected graph, a tree, a flow chart, a circuit diagram, or a state-transition diagram. The input image set can include one image showing the node-link structure or two images, one showing graphical features that are a subset of the nodes and the other an image of an overlay with editing marks that include the links and another subset of the nodes. The input image data are used to obtain likely node-link data indicating parts of the input image set that satisfy a constraint on nodes and parts that satisfy a constraint on links. The likely node-link data are used to obtain constrained node-link data indicating subsets of the likely nodes and links that satisfy a constraint on node-link structures. The likely node-link data can include data defining a likely node image showing parts that meet a node criterion and data defining a likely link image showing parts that meet a link criterion.

Abstract: Input image data define an input image that shows a proportioned parts graph, such as a pie chart or a whole or segmented bar graph. The input image data are used to obtain segmented feature data indicating a feature in the input image that satisfies a constraint on segments. The segmented feature data are used to obtain proportion data indicating each segment's proportion. Various criteria could be applied to find parts of the input image that form the feature. For example, for some pie charts, the segments constraint could include a circularity criterion, a center criterion, and a direction criterion, and proportions could be obtained from directions. For other pie charts, the segments constraint could include a distinct regions criterion and a circularity criterion and proportions could be obtained from directions of region sides.

Abstract: Input image data define an input image set that shows a graphical representation of a layout with two or more segments. The graphical representation can be a sketch, and can include, for example, a rectangular boundary with lines parallel to its sides defining rectangular segments within the boundary. The input image data are used to obtain segment source data indicating a source for each segment and segment position data indicating a position for each segment. The segment source data can indicate, for each segment of the layout, one of a number of source images in the input image set. The segment position data can indicate a reference point and a width and height for each segment.

Abstract: Input image data define an input image that shows a row/column representation, such as a table; a matrix; an array; a calendar; a two-dimensional connection diagram such as a programmable logic array (PLA) diagram; a puzzle or game diagram such as a crossword puzzle, a chess or checkers diagram, a go game diagram, or ticktacktoe. The input image data are used to obtain row data and column data. The row data indicate parts of the input image within a row of the row/column representation. The column data indicate parts of the input image within a column of the row/column representation. The row data and column data are used to obtain content data indicating information represented by the row/column representation for the combination of the row and the column. The content data can be used to obtain output image data defining an output image that includes a table or other representation of the information represented by the row/column representation for each row/column combination.

Abstract: Near neighbor data is produced hierarchically for each pixel of an image. Each pixel's near neighbor data indicates the presence of an approximate near neighbor pixel in each of four quadrants with respect to the pixel, and can also indicate distance and orientation to the near neighbor pixel. The near neighbor data can be produced by parallel processing units, with each processing unit producing, for a respective pixel, a near neighbor data item at every level of the hierarchy. Each near neighbor data item can indicate the presence of an approximate near neighbor in a respective region of the image. A near neighbor data item for a higher level region can be produced by selecting between near neighbor pixels indicated by near neighbor data items for subregions at the next lower level, by choosing the near neighbor pixel that is nearer. The quadrants can be asymmetric so that each pixel is not included in any of its four quadrants and so that the quadrants do not overlap.

Abstract: A body of data is operated upon hierarchically in such a way that, at one or more levels of the hierarchy, the number of aggregative data items produced is not substantially less than the number produced at the preceding level. The body of data can be an image, so that each aggregative data item indicates an attribute of a distinct image region. Such attributes include presence of a single connected component or properties of a component such as width, orientation and curvature. A class of abstract computation structures, called exhaustive hierarchical structures, is introduced in which such dense or exhaustive hierarchical aggregative data analysis processes can be embedded. The embedding of exhaustive hierarchical analysis in a computation structure of this class is analogous, and in some implementations similar in processing efficiency, to the embedding of conventional hierarchical aggregative data analysis processes in tree structures.

Abstract: Shapes are compared by obtaining a measure of dissimilarity based on near neighbor data. First, an image of one shape is analyzed to obtain, for each pixel position, a distance to the nearest pixel that meets a criterion, such as a black pixel. Then, a sum of distances is obtained by totalling the distances of the pixel positions at which an image of another shape meets the criterion. The sum indicates dissimilarity. The near neighbor distances and sum can be obtained by hierarchical operations.

Abstract: Pixels within an image are labeled based on near neighbor attributes. The attributes may be attributes for a near neighbor link or for a near neighbor pixel to which the link leads. The near neighbor attribute data for a pixel may indicate attributes for plural near neighbor links, and a link relationship criterion may be applied to determine whether the links are related in a manner that meets the criterion. For example, a pixel may be labeled only if its near neighbor links include a pair with relative orientation difference of approximately 180.degree., both of which have been previously labeled or are links to a boundary; this criterion is useful in labeling a partially bounded region. A distance independent connected component labeling criterion may be applied during parallel labeling of pixels for rapid connected component labeling. Labeling based on a near neighbor labeling criterion may be iterated until a component or region is completely labeled.