Abstract: A method for identifying fingerprints includes the steps of acquiring a primary image and a secondary image; determining notable points in the primary image; comparing with one another the primary image and the secondary image in order to identify the correspondences between the primary image and the secondary image; and validating the possible correspondences. The comparison between the primary image and the secondary image is based on comparison of the regions which surround the notable points on the primary image, with all the points of the secondary image, through a flash cell array, such as to obtain lists of points in the secondary image which are probably associated with the notable points.

Abstract: A laser range finder (LRF) is used to identify an aircraft approaching a gate. The LRF is directed at the aircraft, and from the echoes, a profile is derived and compared to known profiles. To distinguish among aircraft with similar profiles, the LRF is directed at a volume in which a feature such as an engine is expected and at another volume in which the engine is not expected. The echoes from the two volumes are used to determine whether the engine is in its expected location. If so, the aircraft is identified as the correct type and is allowed to dock at the gate. Otherwise, the aircraft is stopped. The nose height can be used as yet another identifying criterion.

Abstract: The invention provides a method of locating an object within a client region, where the object has a plurality of portions. The method includes defining a full geometric model of the object; defining a plurality of geometric sub-models of the object; determining coarse candidate poses of the object within the region using an image of the region and the full geometric model of the object; determining the fine pose of each of a plurality of portions of the object using the coarse candidate poses of the object within the region, and at least some of said geometric sub-models, to provide a plurality of object portion fine poses within the region; and then determining the fine poses of the object within the region using said plurality of object portion fine poses of a plurality of portions of the object includes determining the coarse poses of each of a plurality of portions of the object, and using each coarse pose, determining the fine pose of each of the plurality of portions of the object.

Abstract: In a method for matching two images, each consisting of a plurality of pixels and having partially overlapping contents, the degree of correspondence between the contents of the images is determined for different displacement positions representing different overlappings of the images. More specifically, a plurality of numbers are determined for each one of a plurality of said displacement positions. Each number is formed with the aid of pixel values from both images and is used to simultaneously retrieve predefined overlap assessment values for at least two of said displacement positions. The overlap assessment values retrieved are subsequently used to determine the degree of correspondence between the images for the different displacement positions. The method is carried out with the aid of a computer and can be implemented as a computer program.

Abstract: In an image recognition method of recognizing a plurality of images from an input image, the plurality of images to be recognized including at least one of a specific image and images provided by rotating the specific image, the method includes the steps of: preparing at least one of mask areas on a window area for each image to be recognized, the window area being structured by cells which are arrayed in two-dimensionally, the cell being formed of at least one pixel, the mask area being structured by at least one of the cells and corresponding to an area representing a feature of the image to be recognized; first determining whether or not the image to be recognized exists based on the result of subjecting pixel density values of the input image in the mask area to operation for each image to be recognized; and second determining whether or not any of the plurality of images to be recognized exists in the window area based on the determination result of the first determining step.

Abstract: A pseudo-inverse derived model-point-to-image-point transformation hypothesis generator as could be used in a computer vision system. Preferred embodiments include generation of n-point transformations directly from n point pair sets and generation of n-point transformations from sets of n point pairs defined by a preliminary three-point transformation plus sets of (n−3) points having minimal error with respect to the three-point trnasformation. Systems include salient corner extraction and hypothesis transformation verification by further sets of point pairs.

Abstract: An image recognition apparatus for recognizing a specific image in an object image using image signals of the object image, the specific image having a possibility of being rotated. The image recognition apparatus includes an image information extraction unit for extracting an object image from the image signals as a candidate for recognition of the specific image and for extracting feature values from the object image using the image signals, and a match determining device coupled to the image information extraction unit for recognizing the specific image in the object image by a comparison of the extracted feature values from the object image with reference data for the specific image. The match determining device is controlled using the reference data according to an easy to determine rotation characteristic and difficult to determine rotation characteristic in the reference data for the specific image.

Abstract: A template is analyzed to determine step sizes for searching within a search area. The template is analyzed by first padding the template with data points to increase its size. Cross-correlation between the padded template and the original template leads to identification of an effective step size along multiple axes. Step sizes for each of a horizontal, vertical and a third axis are derived. Third axis step sizes may correspond to rotation, scaling factor, subsampling factor, linear distance, time or frequency. Windows of the search area, selected based on the step sizes, then are tested in a fast search by correlating the template to selected windows to derive correlation coefficients. Any tested window which has a correlation coefficient exceeding a given value is a potential match for the template and is subject to a refined stage of comparison.

Abstract: A position detection system which performs rough detection processing of detecting a rough position of an approximate position of a check object in a check range based on the comparison result of comparison processing of comparing images provided by compressing the check image and a correct solution image subjected to the detection displacement operation of displacing by rough detection displacement amount at a time in the check image, then performs precise detection processing of detecting a more precise position of the check object in the check range in the proximity of the rough position based on the comparison result of comparison processing of comparing images provided by compressing the check image and the correct solution image subjected to the detection displacement operation of displacing by rough detection displacement amount at a time in the check image at a compression rate more moderate than the previous compression rate, and detects the position of the check object in the check range in multiple

Abstract: Provided is the ability to validate detected features in acquired images to thereby enhance the integrity of any analysis carried out on the detected and validated features. A sequence of images of, e.g., an object is acquired, each image in the sequence corresponding to a distinct orientation of the object about a selected object axis. Images in the sequence are inspected for feature points of the selected feature plane, as-projected into the images, at a first feature detection location and at a second feature detection location. The second feature detection location is configured at an image position at which a feature point detected in the first feature detection location in a first inspected image is expected to appear in a second inspected image. Valid object feature points are identified as being those feature points which are detected in both the first feature detection location in a first inspected image and in the second feature detection location in a second inspected image of the image sequence.

Abstract: Provided is the ability to produce an orthogonal-view representation of a selected feature plane of a three-dimensional object. A plurality of images of the object are acquired, each corresponding to a distinct orientation of the object about a selected object axis. In at least one acquired image, feature points are identified in the selected feature plane as-projected into that acquired image. Feature points are associated with an orthogonal-view representation of the selected feature plane, and feature points from at least one acquired image are correlated with physical orientations on the selected feature plane based on the object orientation corresponding to that acquired image. An orthogonal-view representation of the selected object feature plane can be analyzed for a specified feature configuration even when the acquired object images are perspective-view images. This can be accomplished even through only a subset of feature points may be available in any one given image of the object.

Abstract: An image processing method and apparatus in the form of a sampling mask that enables slight image variations to be absorbed and improves the accuracy of image matching based on a comparison of characteristic quantities such as by the template matching method. A multi-dimensional image is divided into regions that overlap each other. The pixels in each of the regions are weighted in accordance with a predetermined function whereby the weighting is largest at the center of a region and decreases as the distance from the center increases. The characteristic sums of the weighted pixel values are extracted as the mask output for absorbing local image variation.

Abstract: For guiding an airplane to a docking station in an airport, a two-dimensional thermal IR image of a station approach zone is formed from a determined point close to said docking station, Entrance of an airplane into the approach zone and locations of IR features of the landing gear of the airplane are detected on arrival of an airplane. The distance between the airplane and the station is computed from stored characteristics of the landing gear of the airplane expected at the station and from the locations as well as any difference between a current path followed by the airplane and a set path for the airplane to arrive at the docking station in a determined orientation. A system for implementing the method and displaying instructions for the aircraft pilot is also disclosed.

Abstract: A neural network architecture is provided for optical character recognition from an input image in which the target character may be rotated in the image plane. The architecture includes hidden units whose inputs receive image information from portions of the image which are rotationally distributed. That is, the local link between input units and the hidden units is adequate for rotation of the character in the image. Therefore, regardless of the orientation of the image, it is right side up, or approximately so, with respect to one of the hidden units. The hidden units have corresponding inputs with corresponding weight factors, i.e., symmetric weight sharing. Thus, regardless of the orientation of the image, one of the hidden units will produce a high output value indicative of an upright character. Alternatively, a single hidden unit has groups of inputs, each group having a corresponding set of weight factors. The groups are coupled to input units for rotationally distributed portions of the image.

Abstract: A command identifying circuit separates entered image information into a bitmap command representing a line image and an image command representing a half-tone image. A data separating circuit separates layout information, bitmap data and paint-tone data from the separated bitmap command and, in accordance with the layout information, stores the bitmap data in a bitmap memory and the paint-tone data in a block-tone memory. Another data separating circuit separates layout information and image data from the separated image command, and a character/line-image extracting circuit extracts pixels, which constitute a character or line image, contained in the image data separated by the data separating circuit, and stores positional information and tone data regarding the extracted pixels in the bitmap memory and block-tone memory. However, pixels constituting a character or line image are not extracted in an image area within which a line-image area and a half-tone image area overlap each other.

Abstract: One dimensional horizontal and vertical reference projections are obtained for a reference character template. Similar one dimensional horizontal and vertical pattern projections are obtained for an unknown character pattern. The horizontal reference projection and horizontal pattern projection are compared, and the vertical reference projection and vertical pattern projection are compared. The reference character template is then positioned with respect to the unknown character pattern based on such comparisons in order to calculate a correlation coefficient between the two. In certain cases, the unknown character can be identified based solely on its one dimensional projections. Multiple characters in a character pattern are identified either one at a time or as a group. In the latter case, all possible combinations are generated using tree search processing, and the best combination is selected.

Abstract: The invention provides methods and apparatus for processing an image to identify the position of a linear pattern--for example, a line or a cross-hair comprising a plurality of intersecting lines. The system performs a first processing step for generating a projection of the image along axes aligned with an expected position of the linear patterns. A second processing step performs a mirror symmetry filtering on the projection to bring out a single peak corresponding to the center of the linear pattern. To further isolate that peak, the system performs a further filtering operation to remove peaks of lesser slope angle, so that only a highly sloped spike corresponding to the linear pattern will remain. The position of the center that peak corresponds to the center of the linear pattern in the original input signal.

Abstract: A system to align oriented (unaligned) objects in an image to permit such operations as dimensional gauging or measurements to be performed. Using the angle of rotation of the object in the unaligned image, the system creates a destination image having rows and columns of pixels aligned with a specified axis. The unaligned image is then aligned by finding a first pixel location in a first column in the destination image; locating a first whole pixel location nearest to the first pixel area in the unaligned image; copying the actual pixel value of the first whole pixel location in the unaligned image to the first pixel location in the destination image; locating a next whole pixel location in the unaligned image; copying the actual pixel value of the next whole pixel location to a next pixel location in the destination image until all pixels in a first column of the window have been copied to the first column of the destination image; locating the next column, and continuing until a last column is located.

Abstract: A three-dimensional object X is recognized by a method which includes the steps of: a) storing in a memory, a set of two-dimensional enrollment images X.sub.1 -X.sub.n of the object X as viewed from n different directions; b) acquiring a two-dimensional access image of an unidentified object as viewed from a new direction which lies between the n different directions; c) generating, from the stored set of images, a rotated two-dimensional image of the object X as viewed from a direction which is closer to the new direction than any of the n different directions; and, d) identifying the unidentified object as the object X only if a correlation between the image of the unidentified object and the rotated image exceeds a predetermined threshold. By generating the rotated image, misalignments with the access image of the unidentified object are reduced; and thus the predetermined threshold can be raised to reduce recognition errors.

Abstract: A system for detecting, identifying and docking aircraft using laser pulses to obtain a profile of an object in the distance. The system initially scans the area in front of the gate until it locates and identifies an object. Once the identity of the object is known, the system tracks the object. By using the information from the profile, the system can in real time display the type of airplane, the distance from the stopping point and the lateral position of the airplane.

Abstract: A method of determining an indicator for the presence of an image is disclosed in which the image is comprised of a two dimensional array of pixels. The method finds particular use in the prevention of fraudulent copying of images, such as bank notes using colour laser copiers and printers. The method firstly imposes a grid of cells over a portion of the image. Then a representative value for each of the cells is determined. Next, a measure of the difference between the representative values of adjacent cells of the grid is determined over the whole of the grid. The first three steps are then repeated for substantially all possible positions of the grid of cells to determine a grid position having an overall maximum positional difference. Finally, the grid position and the representative value for each of said cells is used as a representation of the image. The representation can thus comprise a template used in an image detection system.

Abstract: The present invention provides an image matching method which allows non-ear image positional shifts to be corrected on a real time basis, and an economical product testing system using this method. A two dimensional test image, consisting of pixels arranged as a matrix of bits distributed in both the vertical direction and the horizontal scanning direction, and an associated reference image are subjected to template matching at a cell matching arithmetic unit for each of line blocks consisting of a plurality of scan lines. The result of the template matching are quantitatively evaluated at an evaluation unit as being good or bad, and positional shifts are sequentially computed according to good evaluation results. Based on the positional shifts thus obtained, the comparative pixel addresses are accordingly corrected before the two images are compared at the comparator.

Abstract: A system is provided for rapidly recognizing hand gestures for the control of computer graphics, in which image moment calculations are utilized to determine an overall equivalent rectangle corresponding to hand position, orientation and size, with size in one embodiment correlating to the width of the hand. In a further embodiment, a hole generated through the utilization of the touching of the forefinger with the thumb provides a special trigger gesture recognized through the corresponding hole in the binary representation of the hand. In a further embodiment, image moments of images of other objects are detected for controlling or directing onscreen images.

Abstract: An image detection system is disclosed for the detection of an image in an input pixel stream. An image is input as a input pixel stream and accumulated and converted into cells which are formed into cell data portions from regions of a corresponding image of the input pixel stream. A template storage ROM stores at least one template that comprises a multiplicity of cell data entries derived form an desired image to be detected. A comparator matrix is connected to the cell converter and the template ROM for comparing the cell data portions with a multiplicity of cell data entries to derive a template detection signal when a predetermined number of the cell data portions match with the multiplicity of said cell data entries. The system finds particular use in the prevention of fraudulent copying of images, such as bank notes using colour laser copiers and printers.

Abstract: The image correlation method and apparatus correlates or matches a test image with a template. The image correlation apparatus includes an image processor for partitioning the template into a number of labels, for determining the total number of pixels N.sub.T which form the template and for determining the number of pixels N.sub.i which form each of the labels i. The image correlation apparatus also includes comparison means for comparing the test image to the template. The comparison means determines, for each predetermined gray level j, the number of pixels of the test image N.sub.j,i representative of a predetermined gray level j which correspond to a predetermined label i of the template. The comparison means also determines, for each predetermined gray level j, the number of pixels of the test image N.sub.j representative of a predetermined gray level j which correspond to the template.

Abstract: Fractal image coding of single images or image sequences is described, which yields an improvement of the encoding result as well as a reduction of computations, whereineach image is divided into range blocks which do not overlap and jointly comprise the complete image contents,range regions are generated, each comprising a range block which is smaller than the range region concerned, and the rest of the range region is filled up with a predetermined signal value,domain regions are generated having the same size as the range regions and comprise parts of the image contents, preferably also in a transformed form,the image data of the range regions and the domain regions are subjected to a Fourier transform,a conjugate-complex signal is formed from one of the two transformed signals and multiplied by the other signal,this product is normalized at a predetermined amplitude,the normalized signal is subjected to an inverse Fourier transform,for each range region, the maximum value of this retransformed signal is d

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: In order to reduce the size of an apparatus capable of obtaining a result of correlation between images at a high speed, a multiplex selector for shifting addresses of a plurality of memories, a detector for detecting an amount stored in a memory element of one of the plurality of memories and an amount stored in a memory element obtained by shifting an address of the one of the plurality of memories, and a determination unit for obtaining correlation between outputs of the detector for each amount of shift.

Abstract: An apparatus and computer-implemented polygon bounding method generate, in a first step, a binary array representation of a closed contour input object relative to a reference coordinate system. Subsequently, in a single pass through the binary array, a bounding polygon is generated as a sequence of parallel "bands," each representing a pair of opposing sides of the polygon. The polygon bands are oriented at successive generally uniform rotations of the reference coordinate system and are positioned to define the bounds of the closed contour for each rotation angle. The spacing and positioning of the bands can thus be used to define polygon size and position at each angle. In this way, comparisons between a plurality of polygons can be made systematically using defined geometrical criteria.

Abstract: A method for recognition of an object in an image is provided, the method comprising the steps of: capturing in electronic memory multiple model images of the object; selecting from each captured model image multiple model image patches; encoding respective selected model image patches as respective sets of inner products of respective image patch data and a set of two-dimensional functions such that the respective sets of inner products include information useful in computing a pose estimate for the object; capturing in the electronic memory a current image; selecting from the captured current image multiple current image patches; encoding respective selected current image patches as respective sets of inner products of respective image patch data and the set of two-dimensional functions such that the respective sets of inner products include information useful in computing a pose estimate for the object; comparing respective current image patch encodings with respective model image patch encodings; identify

Abstract: Apparatus and methods for generating a set of defective images from a model image. Defect parameters used with the apparatus and methods specify classes of defects including resolution, blur, threshold, speckle, jitter, skew, distortion of width and height, offset from a baseline, and kerning. The parameters further permit specification of a range of sizes of the defects and a distribution of the defects within the range. Within the apparatus, actual values for the defects are generated randomly within the specified ranges and distributions. The model image may be represented by means of pixels or by a format which describes the model image's shape. The user may specify the number and the size of the defective images. The defective images are useful for inferring accurate and efficient image classifiers for use in image recognition devices such as optical character readers.

Abstract: Lower accuracy picture coincidence detecting means gives positional relationship between picture data of higher resolution picture and lower resolution picture as corresponding pixel position predicted with marker position as reference, i.e., initial position. Block matching processing means sets a search range corresponding to a block of an arbitrary range set within the lower resolution picture in accordance with the initial position to globally search, within the search range, block on the higher resolution picture corresponding to the block of the arbitrary range to find out a candidate of coincidence area to further interpolate between respective pixels of lower resolution picture and higher resolution picture to thereby evaluate, with high accuracy, similarity between the set block and block within the search range to detect a picture in which coincidence is obtained.

Abstract: A machine vision system for identifying the locations of bonding pads on an integrated circuit mounted in a lead frame. The system involves rotating the image of the die to produce an image of the die unrotated with respect to the lead frame (the "counter-rotated image"). The counter-rotated image is searched with corner templates of the bonding pad. A set of possible candidate bonding pads is created from the location of corner templates found during the search. The set is then scored utilizing matching criteria. The best candidate is selected from the set based upon the candidates' scores. The location of the bonding pad is then generated from the best candidate's corner template locations. The invention can be beneficially applied to locate bonding pads during wire bonding, as well as other machine vision applications.

Abstract: A data input apparatus includes an image pickup with a zooming function, for converting an original image into an electrical signal, an area judger for judging at least a character area from the image taken by the image pickup and a character recognizer for converting the character image, in the character area judged by the area judger, into a character code. A detector detects errors in the character recognition and a controller controls the zooming function of the image pickup according to an output of the detector thereby effecting the character recognition with an enlarged image when errors increase.

Abstract: A system and methods for controlling print quality of a printed product of a printer, which may comprise an image acquisition unit for acquiring an electronic test image of printing on the printed product after printing by the printer, an image processor for comparing the test image of the printing with a prototype image of desired printing and determining a match between the printing on the test image and the printing on the prototype image, and means for generating an alarm representative of a print quality problem when the match between the printing on the test image and the printing on the prototype image does not satisfy a predetermined condition. The printed product is not required to include any reference marks.

Abstract: An artificial visual apparatus and method for image recognition having a simple adaptive scaling mechanism enables the definition of scale invariant visual icons in a processing area corresponding to the anterior inferotemporal cortex (AIT) in a one-step, value-based decision making process. Icon related activity states resulting from sensory filtering to a fourth stage KL filter corresponding to the V4 area are recognized independent of the scale and position of the item to be recognized within the maximum visual field. The AIT processing area controls the window of attention in the V4 area and confines further processing onto this selected spotlight. The invention presents a biologically plausible method for scale invariant mapping from the V4 stage filter to the AIT processor.

Abstract: An aircraft docking system is installed at a point of convenience within the dock in full view of the approaching aircraft and in view of the pilot, and has a display which is readily seen by the pilot through the cockpit windshield as the aircraft proceeds along a taxiway and during a turning of the aircraft from the taxiway into the dock. Operation of the docking system is based on the use of feature extraction templates of three-dimensional images of candidate aircraft for the dock. An image of the approaching aircraft is obtained by use of ranging circuitry of an optical laser radar. By comparing the image with the rotated and scaled model, the image can be interpreted to show the present location and orientation of the approaching aircraft. As the image is updated by the radar, the aircraft travel path, as represented by successive locations and orientations of the aircraft, is determined and is compared with the required travel path for the specific aircraft.

Abstract: Method and system for increasing the effective resolution at which a lower resolution input pixel image is printed on a higher resolution dot matrix printer such an ink-jet or laser printer are described. Illustratively, a 300-dpi by 300-dpi input pixel image is prepared at the printer for printing at 600-dpi by 600-dpi, with otherwise jagged edges appropriately smoothed to produce high print quality. Each pixel and its predefined neighboring pixels in the input image are compared to stored image templates and when a match is found, the corresponding quad pixel, or super-pixel, to be output is altered from a mere replication of the input pixel, thereby to eliminate jagged edges. Memory space in the printer is reduced by storing templates that represent only a fraction of the problem patterns, the input pixel image being reoriented between comparisons in order to account for all possible cases.

Abstract: Pattern recognition, for instance optical character recognition, is achieved by training a neural network, scanning an image, segmenting the image to detect a pattern, preprocessing the detected pattern, and applying the preprocessed detected pattern to the trained neural network. The preprocessing includes determining a centroid of the pattern and centrally positioning the centroid in a frame containing the pattern. The training of the neural network includes randomly displacing template patterns within frames before applying the template patterns to the neural network.

Abstract: A method of recognizing a character including the steps of reading sample characters and generating an image corresponding to the sample characters. Then calculating density projections of each sample characters. A primitive characteristic vector is then extracted from each density projection. Compression then occurs so that an optimal distinction vector and a corresponding characteristic space can be determined. A character to be recognized is then read and an image corresponding to said character to be recognized is generated. A density projection is then made from the image of the character to be recognized. A primitive characteristic vector is then made from the density projection of the character to be recognized. A secondary characteristic vector is then produced by projecting the primitive characteristic vector of the character to be recognized into the compressed characteristic space.

Abstract: An automated check-out system for retail stores, such as supermarkets, employing digitized video signals and other means to determine and differentiate individual products, and thereby register the data for billing and inventory purposes. The sensors control the flow and positioning of retail items into the housing wherein the light source is regulated and controlled. Within the housing, optical imaging of the retail item is made with the electronic signal being digitized. By use of computerized databases, the digitized signal is compared to the finite set of data for products of approximately the same size, shape, weight and other comparative variables. Further confirmation of product identity is achieved by use of bar code differentiation, multiple image comparisons and, if necessary, optical display to the store personnel operating the cash register station serving multiple check-out lanes within the store. The customer is guided through the transaction by use of a video display screen.