Abstract: A massively parallel digital image data processor provides a large number of processing elements arranged in a two-dimensional matrix form. Relative indexed addressing among the processing elements is provided, whereby image data may be easily accessed by and shared among all processing elements. A single-instruction/multiple-data (SIMD) architecture provides instructions to the processing elements in parallel in accordance with specific application programs therefor. The processing elements use triple-ported register files for their internal memory which may input and output data independently and simultaneously. The processing elements are memory-mapped into the address space of the processor's embedded computer to simplify addressing thereof. All image data is inputted and outputted in pixel format. All image data is transferred, stored and processed in bit-serial format.

Abstract: An optical recognition method and system utilizes a scanner to digitize a complex document comprising regions of text and graphics. A recognition engine, illustratively in the form of a general purpose computer, separates regions of text and graphics and identifies lines of text. Character boundaries are determined in each line of text by forming a shadow of the line of text on the horizontal axis. Each byte from the shadow is utilized to access a table stored in a memory associated with the recognition engine. The value contained in the table provides information as to where boundaries are located in the portion of the line of text corresponding to the byte from the shadow.

Abstract: In an object recognition device for judging the normality of an object to be recognized, a monitor region including a plurality of monitor points and being represented by one of the monitor points is superimposed of an image produced by video signals obtained from the object. Additionally, the variation of luminance information obtained from the monitor region with respect to reference information is monitored. In a case where the position of the monitor region on the video image of the object becomes unsuitable for normality determination, a correction range is set at and around the monitor point and the setting of the monitor region is updated to one of a plurality of the monitor points having luminance information obtained when the monitor region is moved to one of the plurality of monitor points contained in the correction range, on which the judgement of normality is facilitated.

Abstract: A method of obtaining a correlation between a pattern to be identified and a number of reference patterns by irradiating and displaying the pattern to be identified and the reference patterns on an optical display plane, thereby producing a complex optical amplitude distribution. This distribution is Fourier-transformed to produce a first power spectrum corresponding to the light intensity distribution of its Fourier transformation patterns. The patterns are again irradiated based upon the first power spectrum and a second Fourier transform is performed to produce a second power spectrum which is the light intensity distribution of the second Fourier transformation pattern. The area of this distribution is restricted by controlling the transmittance or the reflectance of each of the portions of the display plane.

Abstract: A pattern classifier for character recognition is constructed in accordance with a neural network model. The pattern classifier comprises (2n+1).times.(2n+1) input buffer amplifiers and m output buffer amplifiers. The input buffer amplifiers have an inverted output line and a non-inverted output line which intersect input lines to the output buffers. Synapses are selectively arranged at the intersections of the output and input lines in accordance with predetermined mask patterns used in character recognition. PMOS and NMOS transistors are employed for the synapses.

Abstract: A method and apparatus for speeding and enhancing the "learning" function of a computer configured as a multilayered, feed format artificial neural network using logistic functions as an activation function. The enhanced learning method provides a linear probing method for determining local minima values computed first along the gradient of the weight space and then adjusting the slope and direction of a linear probe line after determining the likelihood that a "ravine" has been encountered in the terrain of the weight space.

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.

Abstract: A character recognition system and method is disclosed which provides for the teaching and recognition of characters. The system includes an apparatus and method for obtaining an image which includes the character, and for sampling the character to create a plurality of sample points. A vector correlation is performed on these sample points which compares the stored points to those of known characters. The system also provides a precedence table for determining the correct order of recognition of any correlation of multiple characters.

Abstract: Disclosed is a method of optical character recognition that first segments a graphical page image into word images. The method obtains a set of features by extracting smaller outlines of the dark regions in the word images, and then further dissecting each of the smaller outlines into small sections called micro-features. Micro-features are simply sections of character outlines, therefore, they can easily be extracted from the outlines of an entire word without any knowledge about character segmentation boundaries. Micro-features are extracted from an outline by finding the local extremities of the outline and then defining a micro-feature between each pair of sequential extremities. Once extracted, the micro-features are compared to micro-features from an ideal character in order to classify a character, and convert it into a character code.

Abstract: An apparatus and method for processing video data for compression/decompression in real-time. The apparatus comprises a plurality of compute modules, in a preferred embodiment, for a total of four compute modules coupled in parallel. Each of the compute modules has a processor, dual port memory, scratch-pad memory, and an arbitration mechanism. A first bus couples the compute modules and a host processor. Lastly, the device comprises a shared memory which is coupled to the host processor and to the compute modules with a second bus. The method handles assigning portions of the image for each of the processors to operate upon.

Abstract: An image is tested to determine whether it is substantially like a plurality of sample images by computing statistical information about the sample images and using that statistical information to analyze the image being tested. The analysis is such that the statistical parameters are standardized so that available tables of the central chi-square distribution function can be used, thereby simplifying the necessary calculations. If desired, images which have been determined by the method of this invention to be good can be used to refine the statistical information used in analyzing subsequent images. When an image is identified as unacceptable by the method of the invention, the data for that image can be decomposed in order to identify the parts of the image which are causing the image to be unacceptable.

Abstract: A method and apparatus for differentiating and extracting text and line graphics in an image. The method provides for the use of morphological operations, preferably at reduced scale, to eliminate vertical rules and lines from an image followed by the elimination of horizontal rules and lines, remaining text regions are then solidified to produce a separation mask. The mask is used in conjunction with the original image to produce separate text and graphics images.

Abstract: This invention relates to an image binarization threshold computing apparatus comprising a window scanning means for setting a window of predetermined area over a multilevel image and scanning the same image, a first information generating means for generating a first data on brightness of the image within the window, a second information generating means for generating a second data on fitness of the image within the window with respect to a preset pattern, a third information generating means for generating a third data serving as a picture quality assessment criterion from the first and second data, and an optimal threshold computing means for calculating the optimal threshold value providing the best picture quality according to the third data.

Abstract: A method and an apparatus for processing the predetermined number fields containing the handwritten numeric entries. The numeric entries are subjected to character recognition analysis, and a confidence level is associated with each such numeric field reflecting the degree of confidence with which the apparatus has recognized the numeric dollar amounts. The transaction is automatically balanced from the data in the system by computing a combined value of the recognized dollar amounts making up a transaction and comparing the combined value with the transaction total. This computation and comparison is performed automatically, without the prior intervention or assistance of a human operator to review, verify, correct or otherwise modify the dollar amounts recognized by the system.

Abstract: A hand-written character recognition apparatus with a personal dictionary preparation function has a character recognition unit for comparing an input hand-written character data with a registered standard dictionary to recognize a most similar character pattern, and a display unit for selecting and displaying a display font corresponding to the recognized character pattern, so that a personal peculiar character or simplified character inputted can be recognized. When a personal dictionary is to be registered, the character to be registered is inputted by hand-writing, the character recognition unit compares the input hand-written character data with the standard dictionary to recognize a similar character pattern, and the recognized character code of the standard dictionary is selected as the character code to be registered.

Abstract: An optical apparatus for a fingerprint identification system having a multi-LED for emitting a light, a prism having a slant surface to which a fingerprint being identified is contacted and in which the incident light is absorbed at each ridge but reflected at each valley of the fingerprint, a cover glass functioning as a partial reflecting glass adapted to partially relfect the beam representing a fingerprint image reflected from the prism, a target adapted for focusing a virtual image of the fingerprint thereon, a magnifying glass for magnifying the fingerprint image partially reflected by the cover glass, thereby allowing the user to easily see the fingerprint image, a producing lens for producing the fingerprint image transmitted through the cover glass, and a CCD camera module for converting the fingerprint image produced by the producing lens into an electric signal.

Abstract: A method for determining an image point in an object image composed of the steps of detecting an image signal made up of a series of image signal components representing respective picture elements in a radiation image, which includes an object image and which has been recorded on a recording medium. The center of gravity of the object on the recording medium is found from the image signal by weighting the respective picture elements with corresponding absolute values of differentiated values resulting from the carrying out of differentiation processing on the image signal values corresponding to the respective picture elements. A position, at which the center of gravity is located, is determined as the image point in the object image.

Abstract: Human beings judge patterns (such as images) by complex mental processes, some of which may not be known, while computing machines extract features. By representing the human judgements with simple measurements and reducing them and the machine extracted features to a common metric space and fitting them by regression, the judgements of human experts rendered on a sample of patterns may be imposed on a pattern population to provide automatic classification.

Abstract: A method for highlighting particular features of an image photographically captured and stored in a computer is disclosed. The method essentially comprises capturing an image of an object using a video or similar electronic camera and storing the image as a series of numbers in the memory of a computer. The computer contains an image capture board and a graphics board so as to acquire and display the image on a monitor. The operator enlarges a region of the image of interest so as to display the individual pixels comprising the image. The operator, through the computer, selects individual pixels from the image to comprise a test set of pixels to be used to discriminate against all of the pixels comprising the image. Once the test set is defined, the eigenvectors and eigenvalues of the test set are calculated to define an ellipsoid encompassing substantially all of the test set data points. In essence, the ellipsoid defines an envelope in color space.