Abstract: An image inquiry apparatus includes a plurality of unit inquiry circuits arranged in a matrix of m rows and r stages (each of m and r is an integer more than 1). The plurality of unit inquiry circuits other than the unit inquiry circuits of i-th row and (i+1)-th stage (1≦i≦ a smaller one of m and r, and i is an integer) are connected in series in units of rows. Inputs of the unit inquiry circuits of i-th rows and (i+1)-th stages are respectively connected to inputs of the unit inquiry circuits of i-th rows and i-th stages. Each of the plurality of unit inquiry circuits includes a plurality of pixel processing circuits connected in series, and outputs a calculation data corresponding to a difference between a reference pixel and a retrieval pixel which are both held therein. R summing circuits are respectively provided for the r stages of the matrix.
Abstract: There is described a method for determining the relative location between a rectangle and a polygon. First, a box that bounds the polygon is created. An outside indicator is returned if the rectangle is outside the box. An intersect indicator is returned if the box is entirely inside the rectangle. If an odd/even fill process is in use then first the intersect indicator is returned if the rectangle intersects with the polygon or if a subpath bounding box is entirely inside the rectangle. Finally, if the location of the rectangle has not yet been returned, the location of a single point of the rectangle is returned. If a non-zero winding fill process is in use, first, the intersect indicator is returned if at least one corner of the rectangle is in a different side of the polygon than remaining corners of the rectangle.
Abstract: An improved method of pose determination and tracking does away with conventional segmentation while taking advantage of multi-degree-of-freedom numerical fitting or match filtering as opposed to a syntactic segment or feature oriented combinatorial match. The technique may be used to improve image database query based on object shape descriptors by allowing the user to request images from a database or video sequence which contain a key object described by a geometric description that the user designates or supplies. The approach is also applicable to target or object acquisition and tracking based on the matching of one or a set of object shape data structures.
May 21, 1997
Date of Patent:
January 9, 2001
Cybernet Systems Corporation
Thomas M. Peurach, Peggy Whalen, Douglas Haanpaa, Mike Whalen, Charles J. Jacobus
Abstract: An image processing apparatus including an input unit for inputting multi-valued input image data, a quantization unit for quantizing the input image data input by the input unit into output image data having a number of levels which is smaller than that of the input image data, and a correction unit for correcting a quantization error which occurs during a quantization process of the quantization unit, the correction unit correcting the quantization error by diffusing an error amount into input image data which have not yet been quantized by the quantization unit, wherein the correction unit periodically varies the error amount to be diffused into the input image data within a range of values up to and including the quantization error.
Abstract: A system and method for rapidly obtaining information about the suitability of terrain to meet concealment or visibility objectives, such as might be needed for military planning or for determining placement of relay antennas in a cellular communication system, is disclosed. The disclosed system precomputes heuristic visibility fields for a geographic area, then uses the precomputed heuristic visibility fields to rapidly determine points on a map that will provide the desired characteristics of exposure or concealment in response to a query.
Abstract: A high performance method for the compression of floating point format surface normals and the inverse method for the decompression of those compressed surface normals. Each of the three vector components of the surface normal is compressed by subtracting a constant from the floating point format value, then extracting a predefined field, and finally storing the extracted field. Decompression of the compressed surface normal requires first converting the three stored vector components into floating-point format and then adding a predefined constant to each. Typically the surface normals are of unit length.
Abstract: A new method for real time implementation of rate-distortion optimized coding mode selection is disclosed that can be efficiently applied to H-263-compatible video codecs and other codecs of similar type. A normalized rate-distortion model is used to efficiently compute the rate and the distortion when encoding motion-compensated prediction error signals, instead of performing DCT, quantization and entropy-encoding. A fast algorithm is also disclosed that determines sub-optimal values of coding parameters such as the quantization parameter and the Lagrangian multiplier, .lambda., for the trellis search. Very good rate control and picture quality are achieved, especially when the disclosed techniques are applied in very low bitrate video coding.
Abstract: A reversible eight-element discrete cosine transform coding system which provides transform values near to transform values of the original eight-element discrete cosine transform. In a 4.times.4 matrix transform which appears when an eight-element discrete cosine transform is decomposed in accordance with a fast calculation method, transform coefficients (X1, X7, X3, X5) are separated into (X1, X7) and (X3, X5), which are quantized individually making use of the fact that, if (X1, X7) are determined, then values which can be taken by (X3, X5) are limited. (X1, X7) are quantized with step sizes k1, k7 by linear quantizers to obtain quantization values (Xq1, Xq7). Meanwhile, (X3, X5) are divided into global signals and local signals, and the global signals are quantized with step sizes L3, L5 by linear quantizers while quantization values of the local signals are determined using a table. The quantization values are added by adders to obtain quantization values (Xq3, Xq5) of (X3, X5).
Abstract: A multi-spectral imaging system and method for producing an image is disclosed. More specifically, the imaging system produces an image of vegetation for analysis of crop characteristics, such as nitrogen levels, from an area having vegetation and a non-vegetation background. A light sensing unit detects light reflected at multiple wavelengths. The image is segmented into images at different wavelengths such as at the red, green and near infrared wavelengths. The images are combined into a multi-spectral image and segmented into a vegetation image by eliminating all non-vegetation images by using the images at two wavelengths. The vegetation image is analyzed for nitrogen levels by calculating reflectance values at the green wavelength. The images may be stored for further analysis of crop characteristics.
October 10, 1997
Date of Patent:
December 12, 2000
Monte Andre Dickson, Larry Lee Hendrickson, John F. Reid
Abstract: A method for identifying and extracting text data from a table-cell frame. The method includes the steps of tracing connected components of a document image, tracing white contours within a connected component, defining a frame outline based on the white contours, identifying unattached character data inside the frame outline, and defining an initial rectangular area inside the frame outline.
Abstract: An apparatus and a concomitant method is disclosed for encoding wavelet trees in a wavelet-based coding technique, which addresses both balanced and unbalanced wavelet trees and increases the overall coding efficiency.
Abstract: A coding and/or decoding system for an image including bit-sops of different levels, is provided in an image recoding system which recodes bit-map data including small-level bit-map data and multilevel bit-map data in an image memory and which read the bit-map data out of the image memory to recode an image. In this coding and/or decoding system, after small-level bit-map data and multilevel bit-map data are divided into blocks, each of which is composed of a plurality of picture elements, the divided small-level bit-map data and the divided multilevel bit-map data are inputted to small-level bit-map data coding section and multilevel bit-map data coding section, respectively. The small-level bit-map data are transformed into codes representative of the number of black or white and half-tone picture elements in the block, and the multilevel bit-map data are transformed into codes representative of the average value or the total in the block.
Abstract: To provide a device with facsimile function and printer function whose memory will not be full of facsimile data even if print data priority mode is turned ON while facsimile data is being received. While facsimile data is being received or printed (S11,S12), if print data is inputted (S13), facsimile data print discontinuation flag is turned ON (S15). On the other hand, if no print data is inputted (S13), or if the print data priority flag is in the OFF state (S14), program is terminated without shifting to print data priority mode (S16). When the facsimile data is not being received nor printed (S11,S12), if the print data priority flag is in the ON state (S14), the facsimile data print discontinuation flag is turned ON (S15). Then, the print data priority mode is turned ON. If the print data priority flag is in the OFF state (S14), the program is terminated without shifting to the print data priority mode (S16).
Abstract: In a method for separating a desired color region from a color image, color component values, which represent color components of the image, are calculated. Energy minimization is then carried out, in which calculations of an update rule for minimizing energy are iterated, in accordance with an energy function, which is defined by the color component values and a line process representing the presence or absence of continuity of the color component values in the image. A contour of the color region, which contour is represented by the line process and obtained from the energy minimization, is then extracted. The desired color region is thereby separated accurately from the color image without being affected by a background in the image.
Abstract: A measurement system (1) has robotic system (3) mounted over a base plate (2) which provides a vertical height reference. In the optical head (8), a CCD camera (17) is adjustable to be pre-set in height. The optical head (8) also has an upper light source (18) comprising LEDs and a lower light source (25) comprising a white fluorescent lamp (26) for lateral component illumination. An air flow system (45) directs cooling air in a swirling motion around the tube. An image analysis computer (10) automatically performs various tests to co-ordinate optical and mechanical operation.
August 1, 1997
Date of Patent:
November 21, 2000
MV Research Limited
Peter Conlon, Sean Michael O'Neill, James Mahon
Abstract: A method of coding an object of an arbitrary shape. The arbitrary shape coding method includes the steps of: (a) determining whether or not a shape in a video object layer (VOL) is an arbitrary shape; (b) calculating the size of a video object plane (VOP) in each frame if the shape in the VOL is determined as being an arbitrary shape in the step (a); (c) coding information of each frame about whether or not all macroblocks in the VOP are opaque; (d) coding the VOP in macroblock units if the shape of each frame is determined not to be all opaque in the step (c); and (e) coding only motion and texture information in the VOP, by macroblock, if the shape in the VOL is determined to be a non-arbitrary shape in the step (a) or if the shape in the VOP in a frame is determined as all opaque in step (c).
Abstract: Taking into consideration the disadvantage that a large-scale analog neural network cannot be constructed as an LSI and, even if this were possible, the cost would be prohibitive and the network would lack universality, a digital image processor for processing input image data based upon a cellular neural network is provided with a first multiply-and-accumulate arithmetic unit for digitally processing multiplication and accumulation of input image data of a plurality of pixels and input weighting values in a predetermined area, a second multiply-and-accumulate arithmetic unit for digitally processing multiplication and accumulation of output image data of a plurality of pixels and output weighting values in a predetermined area, and a non-linear acting unit for deciding output image data in accordance with results of calculation from the first and second multiply-and-accumulate arithmetic unit and non-linear characteristic parameters.
Abstract: A pointing device incorporates a biometric sensor at a location such that when operating the pointing device in a normal manner, a user's hand rests naturally in a position to place a finger of the user's hand in proximity to and readable by the biometric sensor. In one embodiment, a computer trackball pointing device includes a fingerprint sensor which is equally well suitable for use by either a right-handed or a left-handed user. Along with positional information from a position sensor and user selection information from at least one user-depressable button, the pointing device also conveys to an attached computer system information associated with the user's identity detected by the fingerprint sensor.
Abstract: In a photogrammetric analytical measurement system, a survey map is produced on the basis of two respective photographed pictures obtained at two different photographing positions. Each of the pictures includes an object image to be recorded on the survey map, and a standard measurement scale image for producing an accurately scaled object image. Data for producing the survey map is approximately calculated on the basis of two-dimensional position data inputted to the computer by designating the scale image and the object image on the pictures with a cursor, by an operator's manipulation of a mouse. When the results of the approximate calculations are erroneous, it is indicated that the inputting of the two-dimensional position data, by designation of the object image with the cursor, should be repeated.
Abstract: An organism identification system that can identify whether or not a fingerprint image input object is a living body at a fingerprint collating time. A finger collator slides his finger at the time of inputting a fingerprint image. Meanwhile the finger is contact with bioelectric potential lead-out portions and a bioelectric potential grounding portion. Sliding of the finger causes a variation in the muscle potential of the finger. The bioelectric potential lead-out portions and the bioelectric potential grounding portion output electric signals corresponding to the potential variation to a bioelectric signal amplification section. The bioelectric signal amplification section amplifies the difference between electric signals from the bioelectric potential lead-out portions while the bioelectric potential grounding portion is at the ground potential. A signal peculiar to a living body is generated by sliding the finger. When an illegal person slides a replica, a variationless signal occurs.