Abstract: A method and computer program product to create an unlimited number of synthetic but realistic, biologically-based, 2-D images like irises and magnetic resonance images (MRIs), as well as other images is presented. New metrics for measuring the mathematical distance of such synthetic images from a source (original) image have also been proposed. These metrics and the synthesis procedure are applicable to the development of image retrieval systems. The presented method can be extended to synthetic images of non-biological origins too.

Abstract: A cursive character handwriting recognition system includes image processing means for processing an image of a handwritten word of one or more characters and classification means for determining an optimal string of one or more characters as composing the imaged word. The processing means segments the characters such that each character is made up of one or more segments and determines a sequence of the segments using an over-segmentation-relabeling algorithm. The system also includes feature extraction means for deriving a feature vector to represent feature information of one segment or a combination of several consecutive segments. The over-segmentation-relabeling algorithm places certain segments considered as diacritics or small segments so as to immediately precede or follow a segment of the associated main character body.

Abstract: A method and computer program product to create an unlimited number of synthetic but realistic, biologically-based, 2-D images like irises and magnetic resonance images (MRIs), as well as other images is presented. New metrics for measuring the mathematical distance of such synthetic images from a source (original) image have also been proposed. These metrics and the synthesis procedure are applicable to the development of image retrieval systems. The presented method can be extended to synthetic images of non-biological origins too.

Abstract: A cursive character handwriting recognition system includes image processing means for processing an image of a handwritten word of one or more characters and classification means for determining an optimal string of one or more characters as composing the imaged word. The processing means segments the characters such that each character is made up of one or more segments and determines a sequence of the segments using an over-segmentation-relabeling algorithm. The system also includes feature extraction means for deriving a feature vector to represent feature information of one segment or a combination of several consecutive segments. The over-segmentation-relabeling algorithm places certain segments considered as diacritics or small segments so as to immediately precede or follow a segment of the associated main character body.

Abstract: Blocks in frames of a video are matched to similar blocks across frames using qualitative characteristics of the pixels in the block. Pixels in a frame are categorized into two or more categories. In one embodiment, three pixel categories are: 1) quasi-constant (QC) pixels having intensity values that vary slowly (e.g., pictorial portions), 2) textured pixels characterized by many small edges and thin-line signals (e.g., texture portions), and 3) dominant-edge (DE) pixels characterized by few sharp and dominant edges (e.g., edge portions). Blocks in different frames are matched by minimizing a cost function that measures the mismatch between a block in a first frame and a plurality of candidate blocks in a second frame. In one embodiment, a cost function comprises a contrast component and a brightness component. The brightness component is determined as a function of the pixel categories. Block matching in sequential frames of a video enables a calculation of displacement or motion estimation in the video.

Abstract: A system and method for classifying a speech signal within a likely speech signal class of a plurality of speech signal classes are provided. Stochastic models include a plurality of states having state transitions and output probabilities to generate state sequences which model evolutionary characteristics and durational variability of a speech signal. The method includes extracting a frame sequence, and determining a state sequence for each stochastic model with each state sequence having full state segmentation. Representative frames are determined to provide speech signal time normalization. A likely speech signal class is determined from a neural network having a plurality of inputs receiving the representative frames and a plurality of outputs corresponding to the plurality of speech signal classes. An output signal is generated based on the likely stochastic model.

Abstract: Pixels in a low bit-rate encoded image are categorized into two or more categories. In one embodiment three pixel categories are used: 1) pixels belonging to quasi-constant (QC) regions where the pixel intensity values vary slowly (e.g., pictorial portions), 2) pixels belonging to textured regions which are characterized by many small edges and thin-line signals (e.g., text portions), and 3) pixels belonging to dominant-edge (DE) regions which are characterized by few sharp and dominant edges (e.g., edge portions). In one embodiment DE pixels are categorized first, then QC and textural pixels are distinguished from the remaining pixels using the number of zero-crossings among the pixels and a threshold. Conventional spatial filters that are well suited for each pixel category type are then used to enhance each region of the image. In one embodiment, various combinations of spatial filters are used to enhance the image.

Abstract: A system for enhancing the quality of an image that suffers from staircase effects, contouring effects, and granular noise, without affecting the existing method of creating the image is disclosed. The enhancement system identifies the edge and nonedge regions in the decoded image, and then filters those regions in order to reduce the specific type of quality degradation that generally occurs in the specific type of region. The system is integrated into a processing system including means for receiving and displaying digital image data. The system receives the digital image data, differentiates between edge data and the nonedge data within the digital image data, and filters the image data accordingly. The filtered edge data and the filtered nonedge data are combined to form an enhanced image. A multi-pass filtering system is also disclosed.