Abstract: A technique is described for table grid detection and separation during the analysis and recognition of documents containing table contents. The technique includes the steps of table detection, grid separation, and table cell extraction. The technique is characterized by the steps of detecting the grid lines of a table using, for example, inverse cell detection, separating noise and touching text from the grid lines, and extracting the cell contents for OCR recognition.

Abstract: A technique is provided for enhancing the background of an original document image. In the case of a black-and-white image, the background color of the original document image is detected, the desired enhanced background color of the original document image is determined from a background pixel value Pb that is in the center of the background color range, and the original document image is enhanced to the desired enhanced background color. However, if the background of the original document image is in color, the technique further includes obtaining color image histograms of red, blue and green colors of the original document image, smoothing the histograms, and comparing the histograms to determine if they have the same shape.

Abstract: A method, computing device, and associated computer readable storage media containing instructions for binarizing a grayscale image by manually determining a first threshold that yields optimal binarization values to one or more images in a set of images, calculating the histograms of each of the images determined using the first threshold, calculating a set of statistical parameters such as the mean, standard deviation and variance of each histogram, determining a second threshold as a function of the set of statistical parameters, and comparing each pixel of the grayscale image to the second threshold. The second threshold T may be a function of the mean m, standard deviation s and variance v and is calculated by fitting a third degree polynomial curve T=a0+a1m+a2s+a3v, where the coefficients A=[a0 a1 a2 a3]T are found using a minimum mean square error algorithm.

Abstract: A method, computer readable medium, and device for reducing speckle in an image by detecting the edges of the image to create an edge detected image, binarizing the edge detected image to create a binary edge image for processing, creating a list, L, of connected components in the binary edge image, creating a list, C, of connected components in list L that are smaller than a predetermined number of pixels, determining noise candidate pixels from the edge detected image that are covered by the connected components in list C, computing a histogram he of the noise candidate pixels, calculating a threshold from the total number of noise candidate pixels, and marking the pixels in the connected components in list C having a pixel intensity smaller than the threshold as noise. The pixels marked as noise may then be removed by setting the pixels marked as noise to a background color of the image.

Abstract: A method, device and computer readable storage media for enhancing an image for optical character recognition by detecting the edges of the image to create an edge detected image, binarizing the edge detected image to create a binary edge image for processing, dilating the binary edge image to create a dilated binary edge image, taking the XOR difference between the binary edge image and the dilated binary edge image to obtain a text boundary, superimposing the text boundary on the image and determining the pixels of the image that are covered by the text boundary, calculating the average grayscale value of the pixels of the image that are covered by the text boundary, and setting background pixels of the image to the calculated average grayscale value of the pixels of the image that are covered by the text boundary.

Abstract: A method and corresponding computing device and computer readable storage media containing instructions for modifying the histogram of a grayscale image to improve contrast by extracting black connected components from the grayscale image that touch at least one of the margins of the grayscale image, computing the histogram of the portion of the grayscale image covered by the extracted black connected components, and updating the histogram of the grayscale image by subtracting the histogram of the portion of the binary image covered by the extracted black connected components from the histogram of the grayscale image or by subtracting a function of number of pixels of the portion of the binary image covered by the extracted black connected components from the histogram of the grayscale image. The function may be a property of a document containing the grayscale image, such as the size of the document.

Abstract: The boundaries of a scanned digital document are determined by identifying the largest connected component in the received digital document and assigning the boundaries of the largest connected component as the boundaries of the received digital document or by using a row by row and column by column analysis of the received digital document to identify horizontal and vertical bands in the digital image having pixels with a value opposite to the value of pixels of a background of the received digital document and assigning the horizontal and vertical bands to be the boundaries of the received digital document. These processes may be performed in series or parallel by a processor associated with a scanner that creates the digital document.

Abstract: A document input device having a gravity feed paper tray is adapted such that the separator pad is mounted on the base, as opposed to the cover, so that when the cover is opened to access the paper path (as during clearing of a paper jam), paper in the paper tray is held in place by the force of the separator pad against the pick roller, even though the cover is open. The separator pad may be attached to a rod that is attached to the base so as to permit the rod and separator pad to remain in place when the cover is in the open position. A leaf spring may be placed between the rod and the separator pad for biasing the separator pad against the pick roller. Also, the cover may include a spring for biasing the separator pad against the pick roller when the cover is in the closed position.

Abstract: A system and method that transfers data from scanned documents and document images directly into a spreadsheet. The user can construct a map that associates data types in the input scanned document with an area in the spreadsheet. The user can also use pre-stored maps that have previously been constructed by the user or by someone else. The map may be stored as an XML file in a hidden sheet of the spreadsheet or in a separate file. During use, the user selects a map, scans the document, parses the document to extract the data types and associated data, and transfers the parsed data to the spreadsheet in accordance with the selected mapping.

Abstract: An automatic document classification system is described that uses lexical and physical features to assign a class ci?C{c1, c2, . . . , ci} to a document d. The primary lexical features are the result of a feature selection method known as Orthogonal Centroid Feature Selection (OCFS). Additional information may be gathered on character type frequencies (digits, letters, and symbols) within d. Physical information is assembled through image analysis to yield physical attributes such as document dimensionality, text alignment, and color distribution. The resulting lexical and physical information is combined into an input vector X and is used to train a supervised neural network to perform the classification.