Abstract: An optical character recognition method and system adjusts positional, rotational and scaling differences between an input form image and a standard form image based upon selected corresponding portions of the two images prior to optically recognizing characters in predetermined areas of the input form. The corresponding portions include a predetermined mark such as a cross-line portions and or predetermined characters, and a selection process of these corresponding portions eliminates unlikely pairs of the corresponding portions based upon a distribution histogram analysis.

Abstract: Spatial locations of individual points on an inaccessible object are determined by measuring two images acquired with one or more cameras which can be moved to a plurality of positions and orientations which are accurately determined relative to the instrument. Once points are located, distances are easily calculated. In distinction to prior art, which uses a fixed separation of camera viewpoints, this new system offers smaller errors, measurement over a larger range of object distances, and measurement of distances which cannot be contained in a single camera view. Random errors are minimized by use of an optimum measurement geometry. Systematic errors are minimized by use of a complete and robust set of calibration procedures. A standard measurement procedure automatically obtains the optimum measurement geometry. A least squares calculation uses all of the image location and calibration data to derive the true three dimensional positions of the selected object points.

Abstract: A method of forming a template of an image of an object includes the steps of: detecting occurrences of at least one feature type within the image which meet a respective criterion; for each such occurrence, determining a feature position which is a point within the image at which the occurrence of the feature type is to be considered as being located; and constructing a structural mesh of links between the feature positions. A method is also disclosed of forming a template derived from a plurality of templates formed from respective images of different members of the class of objects and combining the templates to form a generalized template. In particular the templates are combined using a genetic algorithm. The invention provides a method of forming a template from one or more images which does not rely on a priori assumptions about the salient features of the object for which a template is to be obtained.

Abstract: An object of the present invention is to generate a vector or data from inputted drawing data interactively by suppressing the input cost.When an unknown process occurs in the system during data processing, an inquiry to a user is generated in the case of interactive operation. However, when drawings are handled, the number of inquiries to a user increases and the system operability lowers inversely. The present invention executes the drawing recognition method by cooperation and and competition of a multi-agent on the basis of the basic property of a figure such as adjacent relation. When blurring or a problem portion such as in interruption part or a junction part appears during recognition of a drawing, the linkage relation of line paths is automatically decided by making the recognition agent adjacent to the problem portion cooperate and compete. An inquiry to a user is not made. By doing this, vector data is obtained.The input cost of figure information in a data base can be reduced remarkably.

Abstract: A method for automatic matching of planar point patterns in a reference pattern with a test pattern includes course mating, calculation of a mated possibility for one point in each pattern using fuzzy relaxation, preliminary selection of mated pairs based on these possibilities, and calculation of a similarity value based upon the number of mated pairs, the mated rate of the reference and test patterns, and the average mating possibility and average distance between mated points.

Abstract: A sequential polygon approximation apparatus for a contour and a method thereof which are capable of sequentially extracting a vertex in the order of a contour pixel when performing a polygon approximation so as to code a shape information, which includes an error computation unit for obtaining a contour pixel between two contour pixels of the vertex extraction unit from the contour pixel memory and computing an error which is a distance between a straight line between two contour pixels and a pixel from the contour pixel memory, a comparison unit for comparing an error obtained by the error computation unit with a threshold value, and a controller for controlling the vertex extraction unit and the error computation unit in accordance with the comparison result of the comparison unit, setting the pixel just before the final pixel of a first region as a first vertex when the error obtained in the first region which is the region of the contour pixel sequentially determined for the vertex extraction from a fixe

Abstract: In a template matching method, a region that has a plurality of normalized reference template images as vertex template images in a space of a template-pixel-size dimension is considered, and the vertex template images are linearly interpolated by a parameter w to provide a normalized interpolated template image of a magnitude 1 in the space. Then, matching between the normalized interpolated image and a search image is performed using a normalized correlation therebetween as a similarity measure, by which is obtained a parameter that provides the maximum normalized correlation value, and this normalized correlation value is determined as the maximum one.

Abstract: A design for a high reliability recognition system utilizes two optimized thresholds for each class k of a prototype data base. One threshold is a class region threshold CR.sub.k and the other is a dis-ambiguity threshold DA.sub.k. CR.sub.k specifies a constrained region belonging to a class k, and DA.sub.k corresponds to a value with which a sample belonging to class k can be correctly recognized with a high level of confidence. During recognition, if the distance D(x, r.sub.M) between an input sample x and the representative prototype r.sub.M of a nearest class M is larger than the class region threshold CR.sub.M, x will be rejected. Furthermore, if the distance D(x, r.sub.M) is subtracted from the distance D(x, r.sub.S) between x and the representative prototype r.sub.S of a second nearest class S, the resulting distance difference must be greater than the dis-ambiguity threshold DA.sub.M, or x will be rejected. An inventive algorithm is used to compute optimum thresholds CR.sub.k and DA.sub.

Abstract: Reference patterns and an input pattern are expressed by the position coordinates of feature points obtained by approximating strokes by a predetermined points in advance. One-to-one stroke correspondences whose stroke-numbers are made equal to each other between reference pattern and the input pattern having different stroke-numbers and different stroke-orders each other are determined, and handwriting distortion is stably absorbed by performing an optimum affine transformation operation for each corresponding stroke pair. The inter-pattern distance between a normalized input pattern obtained by the affine transformation operation and the reference pattern is calculated, and the character categories of the reference patterns are sequentially output as candidate categories in order of increasing distance.

Abstract: A method is described for the automatic recognition of at least one handwritten symbol. A representation of such symbol is provided, in a data storage medium, as a sequence of one or more discrete strokes having respective initial and final endpoints, a spatial location and a tangent angle being associated with each endpoint. Representations of plural library symbols, each a prototype of a possible class to which the input symbol may be assigned, are retrieved from a data storage medium. A comparison between the input-symbol representation and each of the retrieved library symbol representations includes comparing the respective tangent angles of the input symbol with the corresponding tangent angles of at least some of the library symbols, and comparing the respective endpoint locations of the input symbol with the corresponding endpoint locations of at least some of the library symbols.

Abstract: A graphic form recognition apparatus which can extract a singular point of a curved line portion precisely is disclosed.

Abstract: A method for generating information regarding a 3D object from at least one 2D projection thereof, the method comprising providing at least one 2D projection of a 3D object, generating an array of numbers described by:.alpha..sub.ijk =v.sub.i 'b.sub.jk -v.sub.j "a.sub.ik (i,j,k=1,2,3),where a.sub.ij and b.sub.jk are elements of matrices A and B respectively and v.sub.i ' and v.sub.i " are elements of vectors v' and v" respectively, wherein the matrices and vectors together describe camera parameters of three views of the 3D object and employing the array to generate information regarding the 3D object.

Abstract: Edge or contour information is extracted from an image array by filtering and encoded. In order to improve reproduction accuracy, two separate filters are used to detect edge information. One of the filters detects "sharp" edges, or edges in which the discontinuity in pixel intensity values occurs over a range of a few pixels. The other filter detects "level" edges in which the pixel intensity value discontinuity occurs over a larger range of pixels than the "sharp" edges. The "smooth" areas between edges or contours are assumed to vary continuously between the contours, but for efficient implementation, a one-dimensional linear interpolation is used to regenerate the contour information between edges. In addition, to further improve performance, a line is fitted to the pixel intensity data. The end values of this line are then used for the pixel intensity data.

Abstract: In order to permit a method of determining the shape of a body, which comprises the step of detecting a contour of said body, to be executed rapidly and with little computing effort, the contour of the body is converted into a contour function which is independent of the position of said body, and said contour function is compared with a predetermined reference contour function for a known shape.

Abstract: A method of processing an image representing an object of substantially uniform intensity contrasted against a background of uniform intensity includes the determination of a trace (.increment.) of the image cutting the object in a segment (I) and the scanning of the image by said trace starting from an origin pixel (yl). Firstly, at each current pixel (Pi) with ordinate (yi), a first function ?E(a,yi), V(a,yi)! of stored cumulated sums is determined, computed beforehand over intervals of the trace which are delimited by the origin pixel (yl) and by end pixels defined by the abscissa (yi) of the current pixel and by a semi-width parameter (a), and a particular optimal value ?E(yi); V(yi)! of the first function of the cumulated sums is estimated within limits (al, am) of the semi-width parameter. Then, over the set of pixels of the trace, a second function of said particular optimal values ?E(yi), V(yi)! is constructed, and an optimal value ?E(yo), V(yo)! of the second function is estimated.

Abstract: There is provided an image processing method whereby a slope at a first position on an outline of character data in input image data is extracted and phase data at a second position on a background of the character data is also extracted. The second position exists in the normal direction of the slope at the first position and is away from the first position by a predetermined distance. The phase data is based on states of black pixels in a plurality of predetermined directions at a target point. The predetermined distance depends on the size of character data.

Abstract: In a figure input method and apparatus, coordinate points are input, a locus made by a plurality of input coordinate points is stored, and the stored locus is displayed. Thereafter, the kind of figure corresponding to the input locus is assigned, and a correct figure corresponding to the locus is drawn based on the assigned kind of figure and stored locus information.

Abstract: Data compression and decompression are performed on a series of data points, by expressing a plurality of data points as an equation or a series of equations. Variable accuracies, and accordingly variable compression ratios, may be provided. Multiple versions of the compressed data, with corresponding multiple accuracy levels, and resulting multiple decompressions may be displayed for comparison. The method and apparatus may be used for storing network statistical data.

Abstract: An improved method of segmentation and feature-based description for verification/authentication of contour images (e.g. handwritten signatures) is described, directed to a system which employs a set of parameters extracted from a dynamic description of a contour image template to perform segmentation of incoming contour images being verified/authenticated by the system. The preferred embodiment would consist of a central processing unit incorporating a high-performance computer, and of many autonomous verification units (e.g. ATM machines) equipped with inexpensive processing devices. In the central processing unit, a contour image template is segmented into strokes according to the segmentation parameters which are computed using selective properties of the mathematical transformation for said contour image template. For illustration purpose, a discrete cosine transform (DCT) is being used.

Abstract: An image processing apparatus is provided which is capable of a high-speed process for obtaining a projection histogram. A digitizing unit subjects a video signal, obtained by acquiring an image of a target mark with a camera, to horizontal scanning and slicing to obtain a binary signal. A window signal output unit, coupled to the digitizing unit, outputs a window signal in accordance with a preset window area. An in-window signal output unit is connected to the outputs of the digital unit and the window signal output unit to perform a logical AND operation on the binary signal and the window signal and outputs an in-window binary signal. A histogram creating unit is coupled to the output of the in-window signal output unit to create a projection histogram by successively counting the high- and low-level states of the in-window binary signal generated as the horizontal scanning progresses.

Abstract: Pattern recognition, for instance optical character recognition, is achieved by forming a skeletal representation (400) of a pattern (300), processing the skeletal diagram (400) to improve representation of curved lines (308) in the pattern (300), representing the processed skeletal diagram (500) by a connectivity matrix (602), and finding a minimum spectral distance between the connectivity matrix (602) and a set of template matrices corresponding to known patterns.

Abstract: From an image memory of a fingerprint picture, gradient vectors of the picture are calculated. And from distribution of gradient vectors in a subregion, ridge direction of the subregion are determined, and confidence of the determined direction is also defined.

Abstract: A method for determining feature points comprises a step for (a) providing directional gradients and a gradient magnitude for each pixel in the video frame, (b) normalizing the directional gradients by dividing the directional gradients with the gradient magnitude, (c) generating a first edge map having the gradient magnitude for each pixel, (d) generating a second edge map having the normalized direction gradients for each pixel, (e) dividing the first edge map into a plurality of blocks of an identical size, (f) providing, for each of the pixels included in each of the blocks, normalized directional gradients for a set of a predetermined number of pixels from the second edge map, (g) obtaining a variance for each of the pixels included in each of the blocks based on the provided normalized directional gradients, (h) determining a feature point for each of the blocks based on the gradient magnitude and variance corresponding to each of the pixels therein.

Abstract: There is disclosed an image processing apparatus which can edit every area of an original image. Such an apparatus has a designating device to designate a desired area in an original, a memory for dividing the outline of the designated image area into a plurality of line segments and storing data defining the plurality of line segments, a forming circuit to form area data defining the image area based on the data defining the plurality of line segments read out of memory, and a processor to execute the processing of every area of the original image on the basis of the area data formed by the forming circuit. The memory stores the data defining a start point, an end point, and a slope of each of the line segments. With the apparatus, not only a rectangle area but also a plurality of complicated non-rectangle areas can be easily edited and processed by a simple program using a memory of a small capacity.

Abstract: A technique involving the standardization of the subject matter of an image and the compression of the image using features of the standardization can provide for a high compression ratio. This technique can be used to improve the quality of a restored image for a predetermined amount of storage space.

Abstract: The identity of an object is determined by obtaining an image of the objective shape (e.g., by a camera), obtaining from that image an optimum objective shape representative of the object, and comparing the optimum objective shape with a known shape model of the object. The shape model is represented as a plurality of nodal points interconnected by line segments. The optimum objective shape is selected from a number of objective shapes each comprised of a row of dots having the same number of nodal points as the shape model. Each objective shape is compared with the shape model on the basis of (i) the proximity of the dots of the objective shape to the line segments of the shape model, (ii) the length of the dot segments with respect to the lengths of respective line segments, and (iii) the inclinations of the dot segments with respect to the inclinations of the line segments. The objective shape having the closest similarity to (i.e.

Abstract: A system for recognizing handwritten words of cursive script by extracting temporal information from static text is described. The temporal information consists of the shape, location and temporal sequence of the segments, or "primitives", constituting the word and is translated from bit maps images of the words into digital information format. By comparing the digital information with formatted digital information for a plurality of known words, as may be stored in computer memory, a list of words having a relatively high probability of matching the static text is generated, and used for word recognition by a process of selecting the most likely match.

Abstract: A 3D model of a subject is acquired. Sensors spatially located on the subject are displayed on the 3D model in their appropriate locations. An operator selects the proper viewing angle and sensors to monitor via input devices. A 3D modeler in a display processor provides an image viewed from the proper angle and identifies the selected sensors. The 3D modeler identifies the data channels which correspond to the selected sensors and provides these to a system control computer which then provides the samples for these data channels. The 3D modeler also identifies a screen location which pertains to the locations of the selected sensors on the model. A postage stamp trace unit receives the samples for the selected data channels and the screen locations pertaining to the selected sensor locations of the image of the model and produces a video signal representing a window having a number of real-time traces each pertaining to a sensor.

Abstract: A character pattern generating apparatus for generating a bit map data from outline data corresponding to a pattern having a curve portion comprises: a discriminating circuit to discriminate a size of pattern data to be output; and a deciding circuit to decide whether the bit map data is generated or not by regarding the curve portion as a straight line from the outline data on the basis of the discrimination of the discriminating circuit. The curve portion is a portion to form a Bezier curve on the basis of the outline data.

Abstract: A system for identifying freehand curve drawings has: an input section for inputting a freehand curve drawing as a series of point data; a fuzzy-curve information generating section for generating fuzzy-curve information from the series of point data by adding input vagueness information to each of the series of points; a segmentation section for dividing the fuzzy curve information into a plurality of segments by evaluating a degree of standstill at each point; a curve-class identifying section for identifying a curve class and a representative curve in the class by measuring a possibility of the curve class for each segment with respect to each of a plurality of curve classes; and a connecting section for connecting representative curves by adjusting positions of the representative curves at each segment identified by the curve-class identifying section.

Abstract: A recognition system for identifying members of an audience, the system including an imaging system which generates an image of the audience; a selector module for selecting a portion of the generated image; a detection means which analyzes the selected image portion to determine whether an image of a person is present; and a recognition module responsive to the detection means for determining whether a detected image of a person identified by the detection means resembles one of a reference set of images of individuals.