Abstract: A new encoder is provided which operates on a database of n-dimensional signal vectors where such signal vectors are restricted to being non-negative. In the establishment of a codebook for this encoder, p signal vectors (of dimension n) are placed in an n.times.p matrix X, and an approximate factorization: X.apprxeq.WV is constructed, where W and V are matrices whose entries are non-negative. Each column of the n.times.r matrix W is characterized as a "feature", such features having been "learned" from the database of n-dimensional signal vectors.

Abstract: A motion vector estimating system comprises a current image storing unit 1, a reference image storing unit 2, a motion vector estimating circuit 3 for estimating a motion of a current image in relation to a reference image, and a global motion vector estimating circuit 10. The global motion vector estimating circuit 10 comprises at least a feature setting part 11 for setting and recording a feature being a partial region distinguishable from other regions of a principle object in a picture, a feature-coordinates extracting part 12 for searching the current image for the feature to extract the coordinates of the feature, and a feature-coordinates storing part 13 for storing the coordinates of the feature.

Abstract: A novel vector quantization encoding system comprises a conventional vector formatter and a novel gross quantizer. The vector formatter generates vectors to be searched in a full code book of vectors. The gross quantizer generates pointers indicative of subsets of vector indices. The vector indices are indicative of groups of vectors to be searched in the full code book that are substantially less in number than the number of vectors in the full code book. A search controller is employed to search only the subset of vectors in the full code book and produce as an output the index of the one vector in the full code book closest to the formatted vector being searched so that the quality of a full code book searched is made faster without degradation of a full code book search.

Abstract: A digital transmitter/receiver communications system transmits audio voice signals over a channel with increased quality for a specified bit rate. The method of encoding takes advantage of spherical symmetry of error vectors associated with encoding Line Spectral Frequency (LSF) coefficients, to reduce the information transmitted. Errors in encoding the LSF coefficient sets, vectors J, are modeled by a number of vectors J.sub.p having all positive components, and a sign vector s indicating the polarity of each component of the vector. Each LSF vector J intended to be transmitted is approximated by a positive vector J.sub.p and a sign vector s. An index I.sub.p of the positive vector J.sub.p and the sign vector corresponding to vector J are transmitted, along with other audio information to a receiver/decoder where the signal is decoded into an audio signal closely representing the original signal intended to be transmitted.

Abstract: When gradation values of an original image are converted, a pixel data read unit reads out a piece of pixel data, and a pixel data conversion unit converts the gradation value of the piece of pixel data. A condition judging unit decides whether a pixel corresponding to the piece of pixel data satisfies a predetermined condition, and a data conversion correction unit corrects by a stochastic process the conversion performed by the pixel data conversion unit when the condition judging unit decides that the piece of pixel data satisfies the predetermined condition.

Abstract: A method and apparatus is described for improving the efficiency of any machine that uses an algorithm that maps to a higher dimensional space in which a given set of vectors is used in a test phase. In particular, reduced set vectors are used. These reduced set vectors are different from the vectors in the set and are determined pursuant to an optimization approach other than the eigenvalue computation used for homogeneous quadratic kernels. An illustrative embodiment is described in the context of a support vector machine (SVM).

Abstract: A method for reducing the RAM requirement for temporarily storing a stream of data blocks in a coding/decoding system of information transferable by blocks, includes the steps of: compressing and coding the data by blocks through a tree search vector quantization (TSVQ); storing TSVQ compressed and coded data in the RAM; and decoding and decompressing in a subsequent reading of the data stored in the RAM the coded and compressed data, thereby reconstituting the stream of digital data blocks.

Abstract: Printing hints can be efficiently compressed in a system using hierarchical compression (HVQ) by first using HVQ to generate compressed codewords, and then adding the printing hint to the codewords to generate an intermediate form of codeword, before sending this intermediate form to a pattern matching lossless compressor for final compression. Even though the same printing hint may be applied to a number of codewords, the overall compression will be affected very little by the addition of the printing hints since the hints are identical over a large number of codewords in series.

Abstract: A system and method for scaling image streams that use motion vectors is disclosed. The system combines an error term with a predicted term in order to produce a display value. The system operates on image components represented in the spatial and frequency domains. The system processes motion vectors in the spatial domain. The motion vectors are scaled. The integral part of the scaled motion vector addresses a framestore. The fractional portion of the scaled motion vector is input to a nonlinear filter which determines the value of image components for a location that does not correspond with an image location in the framestore. The output of the nonlinear filter comprises the predicted terms. Data in the frequency domain is processed more efficiently by reducing the size of a block of data by appropriate filtering. The resulting data is transformed to the spatial domain.

Abstract: Fuzzy rule-based control is applied to video rate control in the present invention. A basic fuzzy logic controlling apparatus using the occupancy as a fuzzy control variable is provided. Other fuzzy logic controlling apparatus is also provided: the video quality-controlled, the scene change based and the quality-monitored. The occupancy-based controlling apparatus is adopted to form a backbone scheme. The accumulative quality measure and the scene change features are then introduced to improve the video quality. The quality-monitored FRC (FRC-QM), which is controlled by both the quality factor and the scene change-based factor, is superior to the other schemes in respect of both the occupancy and the quality.

Abstract: A coding apparatus divides digital image information into pixel blocks each having a size of m.times.n pixels, and subjects the image information to vector quantization in block units. The pixel blocks are orthogonally transformed and the characteristics and sequency components of the pixel blocks are detected. Each pixel block is divided into a plurality of sub-blocks in conformity with the detected sequency components. Respective ones of the plurality of divided and outputted sub-blocks are scalar-quantized into sub-vectors of a predetermined number of bits in conformity with the characteristics of the pixel blocks. Respective ones of the scalar-quantized values of the quantized sub-blocks are vector-quantized in conformity with the characteristics of the pixel blocks, these vector-quantized reproduction vector codes are combined and the result is subjected to further vector quantization.

Abstract: An image is processed taking into account the direction of a predominant structure of the image. Said predominant direction is derived from image information in the image. In particular, the covariance matrix having matrix elements depending on products of differences between pixel-values in separate directions is calculated. The eigenvectors of the covariance matrix correspond with the predominant direction of the image structure and the eigenvalues of the covariance matrix represent the strength of the structure in the image. The covariance matrix is computed locally, i.e. for separate regions in the images so as to take variations of the direction of predominant structures into account.

Abstract: A method for color error diffusion using semi-vector quantization is proposed that simplifies error diffusion calculations. When three color separations are used, the two separations with the most visually perceptive color output are compared with one another to determine their output colors, while the third, least visually perceptive separation, is treated separately. A correction term may be applied to the third separation based on the color output of the first two separations to insure uniformity of ink coverage. This method yields results close to vector quantization and also minimizes fluctuation in the two most visually perceptive output colors. Of course, this method may be used with more than three color separations. For example, when four color separations are used, the least visually perceptive color is processed differently than the three other color separations.

Abstract: An encoding process is configurable to trade calculational complexity for bit rate while maintaining code format and image quality. A software encoder which implements the encoding process monitors bandwidth and processor utilization and when either is overtaxed, adjusts complexity to compensate. If further adjustment is required, encoded image quality is changed. Changing the search process or search window size used during searches for predictive blocks or changing a threshold which is compared to inter-code length before determining whether to calculate an intra-code changes complexity. An intra-coding method subtracts a base which depends on codes representing neighboring blocks from coefficients for a current block to determine a difference which is often small and easily compressed. Additionally, restricting a quadtree structure for one color component to a sub-tree of a quadtree structure for another other color component improves compression efficiency and reduces coding complexity.

Abstract: In order to code a digital signal (IM), the invention combines a breakdown of the digital signal into frequency subbands and a finite state vectorial quantization of the subbands. The digital signal is analysed in a plurality of frequency subbands distributed in accordance with at least two different resolutions, and then the subbands of higher resolution are coded by finite state vectorial quantization as a function of the subbands of lower resolution.

Abstract: A hashing-based vector quantization process partitions a codebook into buckets. A hashing function determines the locations of buckets in the codebook and is selected so that buckets containing similar codevectors start near each other in the codebook. One hashing function forms an index from the most significant bits of components of a vector. During encoding, the hashing function generates an index from an input vector, and a search for a codevector matching the image vector begins in a bucket indicated by the index. In one embodiment, the codebook contains flag fields to indicate the starts of buckets and pointer fields to create linked lists of codevectors which form the buckets. Codevectors are compared to the input vector by determining the mean squared error of the difference between the codevector and the input vector. A search is complete when the search finds a codevector having a difference with a MSE less than a threshold.

Abstract: In order to progressively transmit a digital image (IM), a method for coding the image comprises the division (E1) of the image into vectors (V.sub.m,n), the selection (E2) of a code vector (VC.sub.d) associated with an index (A.sub.d) for each of the image vectors, and the replacement (E2) of the image vectors by the indices associated with the selected code vectors, in order to produce a plane of indices (PI) representing the image. The method comprises the steps of sub-sampling (E3) the plane of indices (PI) into sub-planes of indices (SPI.sub.1 to SPI.sub.F) respectively comprising indices selected from the plane of indices, and ordering (E4) the sub-planes of indices according to a predetermined order before transmitting them.

Abstract: An image data coding system comprises a screen-area determining means for determining the area of a screen reproduced on the basis of an input image data; a code-amount assigning control means for controlling regions to which data on the screen are to be assigned and the code amount assigned to each of the regions, on the basis of the results determined on the area of the reproduced screen; and a coding means for coding the image data signal inputted in accordance with the code amount assigned to each of the regions. In an image data coding and/or decoding system, a coding system performs the two-dimensional orthogonal transform of picture signals of all the pixels with respect to inside blocks and of only picture signals of pixels contained in a content with respect to edge blocks, in accordance with a map signal indicative of the position and shape of the content, and it codes the map signal.

Abstract: A data compression system 200, method, and apparatus 214 employs an encoder 210 optimized to decorrelate and make independent from the original signal 202, the quantization noise produced during signal compression. The proposed system 200, method, and apparatus 214 supports high degrees of signal compression, which in turn leads to lower computational complexity and improved performance. Because the quantization noise produced during signal compression is made independent from and non-orthogonal (i.e., uncorrelated) to the original signal 202, enhanced filtering is achievable, which in turn leads to improvements in the signal-to-noise ratio (SNR) of the decoder 220.

Abstract: The present invention relates to the field of vector quantization of transmitted imagery. In particular, the invention is a method for automatically generating a codebook for use in the transmitter and receiver components of an amplitude-adaptive normalized differential vector quantization system based upon a specified signal-to-noise ratio goal. In accordance with this method, the specified signal-to-noise ratio goal automatically derives the amplitude thresholds and other required data to determine the tradeoff between image quality and data compression for the transmitted imagery.

Abstract: An apparatus for vector-quantizing an image signal having a low transmission rate which includes a four-band linear filter para-unitary (LPPU) filter group for transforming an input image signal into transform coefficients which are represented by 4.times.4 frequency bands. A code modulator differential-pulse-code-modulates a direct current (DC) component among the transform coefficients output from the four-band LPPU filter group. A coefficient extractor extracts a predetermined number of low-frequency alternating current (AC) component coefficients from the transform coefficients output from the four-band LPPU filter group. A classifier determines a corresponding class, from a plurality of predetermined classes, using the low-frequency AC component coefficients extracted by the coefficient extractor, and outputs class information representing the determined class.

Abstract: An image processing system includes an image encoder system and a image decoder system that are coupled together. The image encoder system includes a block decomposer and a block encoder that are coupled together. The block encoder includes a color quantizer and a bitmap construction module. The block decomposer breaks an original image into blocks. Each block is then processed by the block encoder. Specifically, the color quantizer selects some number of base points, or codewords, that serve as reference pixel values, such as colors, from which quantized pixel values are derived. The bitmap construction module then maps each pixel colors to one of the derived quantized colors. The codewords and bitmap are output as encoded image blocks. The decoder system includes a block decoder. The block decoder includes a block type detector, one or more decoder units, and an output selector.

Abstract: A video signal encoding and/or decoding method based on adaptive lattice quantization. An input frame of a video sequence is partitioned into blocks of pixel data called vectors. The set of the vectors is partitioned into blocks called macroblocks. The vectors in each macroblock are subjected to lattice vector quantization which examines the range of vector components in multiple dimensions and adjusts the density of lattice vectors in accordance with the range of vector components. Quantization unit replaces each input vector for a nearest lattice vector in multidimensional space. Each quantized vector in the macroblock is converted into an index of a lattice vector. An overhead information about the range in a macroblock and set of indices with or without subsequent Huffman or Arithmetic coding is transmitted to or stored in the receiver end.

Abstract: A novel vector quantization encoding system comprises a conventional vector formatter and a novel gross quantizer. The vector formatter generates vectors to be searched in a full code book of vectors. The gross quantizer generates pointers indicative of subsets of vector indices. The vector indices are indicative of groups of vectors to be searched in the full code book that are substantially less in number than the number of vectors in the full code book. A search controller is employed to search only the subset of vectors in the full code book and produce as an output the index of the one vector in the full code book closest to the formatted vector being searched so that the quality of a full code book searched is made faster without degradation of a full code book search.

Abstract: A method and apparatus is disclosed for encoding and decoding an input signal representative of video or audio information. A form of the method includes the following steps: processing the input signal to produce a sequence of vectors representative of the input signal; implementing a lattice vector quantization procedure on individual vectors of the sequence to identify a lattice point of a lattice that is closest to the individual vector and to obtain an index value representative of a labelling of the identified lattice point, the lattice vector quantization procedure including a multi-level allocation of groups of reserved indices, and selection of the index value for the individual vector within a group of reserved indices. The index signals representative of the index values are stored and/or transmitted, and then recovered for decoding. The decoding is operative to obtain a recovered sequence of vectors, and the sequence of vectors can be processed to produce an output signal.

Abstract: A method and an apparatus for encoding an image signal. The apparatus includes an acquisition module disposed to receive the image signal. A first processor is coupled to the acquisition module. At least one encoder processor is coupled to the first processor. The at least one encoder processor produces an encoded image signal under control of the first processor. The method includes the steps of converting an input image signal into a predetermined digital format and transferring the digital format image signal to at least one encoder processor. The method further includes the step of applying, at the at least one encoder processor, a hierarchical vector quantization compression algorithm to the digitized image signal. At the next step, a resultant encoded bit stream generated by the application of the algorithm is collected. The method and apparatus of the present invention may be used in conjunction with an ordinary modem to transmit and/or receive audio, video sequences or still images.

Abstract: A signal compression system includes a coder and a decoder. The coder includes an extract unit for extracting an input feature vector from an input signal, a coder memory unit for storing a predesigned vector quantization (VQ) table for the coder such that the coder memory unit uses a set of primary indices to address entries within the pre-designed VQ table, a coder mapping unit for mapping indices from a set of secondary indices to the first set of indices, and a search unit for searching for one index out of the set of secondary indices, wherein the index from the set of secondary indices corresponds to an entry in the coder memory unit, and the entry best represents the input feature vector according to some predetermined criteria.

Abstract: A method for a multidimensional search for a close neighbor of an input vector, amongst a first set of reference vectors, comprises the prior determination of a first set of hyperplanes in the space containing the reference vectors, then selection of a first hyperplane from the first set, formation of a second set of reference vectors, by eliminating reference vectors which are on the other side of the first hyperplane selected, compared with the input vector, formation of a second set of hyperplanes, by eliminating the said first hyperplane, reiteration, a predetermined number of times, of the selection and formation operations, taking, as the first set of reference vectors and as the first set of hyperplanes, respectively, the second sets formed previously, and searching for the closest neighbor of the input vector in the second set of reference vectors.

Abstract: The invention relates to a method and system for allocating bits for representing blocks that are transmitted in an image compression system. In a preferred embodiment, a video telephone communications system includes a video transmitting/receiving (transceiving) system including local and remote processing systems separated by a communications link. The local processing system determines indices to be sent to a remote processing system to represent the image. Under one aspect of the invention, the local processing system determines an optimal bit allocation in which the following relationship may be minimized DR=D+.lambda.R, where D is the total distortion .SIGMA.d.sub.i for a frame; .lambda. is a value determined as described below; R=.SIGMA.r.sub.i .apprxeq.R.sub.d which is the desired total number of bits for the frame. The optimal bit allocation may determine .lambda. for an entire frame. Under other aspects of the invention a lambda feedback technique, with or without predecision, may be used.

Abstract: A method for producing a decoded signal approximate to an original signal upon decoding a quantized signal which is produced by vector-quantizing every block of an original signal or by vector-quantizing every block of a transform coefficient obtained after every block of the original signal is processed by a transform process such as a discrete cosine transform (DCT). The method includes the steps of determining a narrow vector quantization constraint set domain in which each vector or block divided from the decoded signal exists prior to the vector quantization at the transmitting site, and outputting a signal indicative of the determined narrow vector quantization constraint set domain, and performing a projection for returning a vector existing beyond the narrow vector quantization constraint set domain to its original domain, and repeating the projection for returning a vector existing beyond the narrow vector quantization constraint set domain to its original domain.

Abstract: Vectors associated with multiple sources are quantized with a codebook. A mapping function associates each source with a subset of codevectors in the codebook, each subset being defined prior to quantization. Each vector from a source is matched to a codevector in the subset corresponding to the source. The universal codebook is designed by first providing a codebook and a mapping function. Using the codebook, the mapping function is updated by identifying subsets that reduce distortion when vectors from a source are quantized using a corresponding subset.

Abstract: A method is disclosed for encoding and decoding an input signal representative of video or audio information. A form of the encoding includes the following steps: processing the input signal to produce a sequence of vectors representative of the input signal; implementing a lattice vector quantization procedure on individual vectors of the sequence to identify a lattice point of a lattice that is closest to the individual vector and to obtain an index value representative of a labelling of the identified lattice point, the lattice vector quantization procedure including a multi-level allocation of groups of reserved indices, and selection of the index value for the individual vector within a group of reserved indices. The index signals representative of the index values are stored and then recovered for decoding, or transmitted before decoding. The decoding is operative to obtain a recovered sequence of vectors, and the sequence of vectors can be processed to produce an output signal.

Abstract: An encoding apparatus uses a vector quantization encoding method for encoding indexes of codewords, which supply a scalar quantized code of a maximum scalar product value of each code word in a code book, and its maximum scalar product value to a vector component of an input image inputted from an image sensor, so as to output the encoded indexes. A scalar product value calculating circuit in the encoding apparatus has scalar product value calculating sections, which are composed of an analog circuit having a code component capacitor corresponding to each code component, a differential amplifier and a feedback capacitor, corresponding to each codeword, and the scalar product values of the input vectors are calculated in parallel by the scalar product value calculating sections. In such a manner, when the analog calculation is made, the scale of the circuit can be decreased and the power consumption can be lowered.

Abstract: A method of authenticating a video image created by a camera (V) or similar video device. The image is formed into a first 2-dimensional pixel array (A1) with each pixel (p.sub.m,n) represented by a data word of a predetermined length. This formatted array is converted into a second 2-dimensional array (A2) which may be made smaller than the first array by eliminating rows and columns from the formatted array. A first linear vector (A3) is created using the data words in the second array, and a second linear vector (A4) is created by repositioning the data words from the first linear vector in a random pattern. A checksum is created by summing the contents of all of the data words in the second linear vector beginning at a location established by a pre-established formula. A header (H) is formed using the resulting checksum, information identifying the device used to create the image, and the time the image is formed.

Abstract: A system and method for compressing related data sets of a sequence are disclosed. The process compares domain blocks from a current frame buffer to range blocks in a previous frame buffer or vector quantization ("VQ") tables to generate affine map codes. The affine map codes are preferably encoded by an entropy encoder before being transmitted to a remote site for decompression. The cost for the encoded affine map codes are computed and used to determine whether affine map codes for representing smaller blocks should be included in the affine map codes which represent the domain blocks into which the original frame was segmented. The methods are preferably implemented on a commercially available discrete cosine transform ("DCT") processor having a process controller and a data comparator. The results of the affine map code generating process on the DCT processor achieves a more consistent bit rate and image quality than methods operating the DCT processor to generate DCT codes.

Abstract: To transmit and record an input pattern efficiently by dividing the input pattern into a vector quantization index and its difference signal, and sequentially optimize the vector quantization code book. A least distortion vector determining unit 101 issues a representative vector y.sub.m having least square error in y.sub.i in the code book, with respect to an inputted 64-dimensional vector x, and simultaneously issues its index m, and the difference vector d obtained as d=x-y.sub.m in a subtractor 103 is quantized in a quantizing unit 104, and is issued together with the index data. Afterwards, a difference vector d' is determined in an inverse quantizing unit 106, and a code book changing unit 107 changes so that the representative vector Y.sub.m may be closer to the input vector x by using this difference vector. At the decoding side, synchronously, the representative vector y.sub.m is changed.

Abstract: An image compressing apparatus employs a mean-separated normalized vector quantization method according to which, with respect to vector components corresponding to input images inputted from image sensors via a plurality of lines, encodes and outputs a scalar-quantized code of a mean value, a scalar-quantized code of a maximum scalar product value with each code word in a code book, and an index of one of the code words which yields a maximum scalar product value. In this image compressing apparatus, when the maximum scalar product value is less than a predetermined threshold value, in accordance with judgement by a comparator circuit, an output selecting circuit stops outputting the codes of the maximum scalar product value and of the index, and outputs only the code of the mean value. Therefore, when the image is uniform with pixels varying little in their luminance levels in compression processing unit blocks, code data to be outputted are restricted so that only data of the mean value are outputted.

Abstract: A hashing-based vector quantization process partitions a codebook into buckets. A hashing function determines the locations of buckets in the codebook and is selected so that buckets containing similar codevectors start near each other in the codebook. One hashing function forms an index from the most significant bits of components of a vector. During encoding, the hashing function generates an index from an input vector, and a search for a codevector matching the image vector begins in a bucket indicated by the index. In one embodiment, the codebook contains flag fields to indicate the starts of buckets and pointer fields to create linked lists of codevectors which form the buckets. Codevectors are compared to the input vector by determining the mean squared error of the difference between the codevector and the input vector. A search is complete when the search finds a codevector having a difference with a MSE less than a threshold.

Abstract: A method and an apparatus for encoding an image signal. The apparatus includes an acquisition module disposed to receive the image signal. A first processor is coupled to the acquisition module. At least one encoder processor is coupled to the first processor. The at least one encoder processor produces an encoded image signal under control of the first processor. The method includes the steps of converting an input image signal into a predetermined digital format and transferring the digital format image signal to at least one encoder processor. The method further includes the step of applying, at the at least one encoder processor, a hierarchical vector quantization compression algorithm to the digitized image signal. At the next step, a resultant encoded bit stream generated by the application of the algorithm is collected. The method and apparatus of the present invention may be used in conjunction with an ordinary modem to transmit and/or receive audio, video sequences or still images.

Abstract: The apparatus and method presented use a localized model-based approach to compress video image signals for applications such as video teleconferencing. Preferably, the apparatus and method use three-dimensional vector quantization (3D VQ), recent localized history of the image signals, and a pyramidal hierarchial encoding technique to compress the image signals.

Abstract: Image recognition, extraction, and data compression is carried out on the basis of colors. There are provided a judgment section 14 for judging a color difference between a color of each pixel constituting an image and each element of a color table; an image replacement section 16 for replacing, in accordance with the result of judgment, the color of each pixel with an element, among elements of the color table, which minimizes the color difference relative to the color of the each pixel; and a pixel integration section for integrating the adjoining pixels after the replacement, if colors of pixels positionally adjoining in an original image are contained in regions adjoining in a divided color space. A single color replaced with an element of the color table is imparted to a group of pixels integrated by the pixel integration section 18.

Abstract: Improved method and apparatus for vector quantization (VQ) to build a codebook for the compression of data. The codebook or "tree" is initialized by establishing N initial nodes and creating the remainder of the codebook as a binary codebook. Children entries are split upon determination of various attributes, such as maximum distortion, population, etc. Vectors obtained from the data are associated with the children nodes, and then representative children entries are recalculated. This splitting/reassociation continues iteratively until a difference in error associated with the previous children and current children becomes less than a threshold. This splitting and reassociating process continues until the maximum number of terminal nodes is created in the tree, a total error or distortion threshold has been reached or some other criterion. The data may then be transmitted as a compressed bitstream comprising a codebook and indices referencing the codebook.

Abstract: A method and an apparatus for encoding an image signal. The apparatus includes an acquisition module disposed to receive the image signal. A first processor is coupled to the acquisition module. At least one encoder processor is coupled to the first processor. The at least one encoder processor produces an encoded image signal under control of the first processor. The method includes the steps of converting an input image signal into a predetermined digital format and transferring the digital format image signal to at least one encoder processor. The method further includes the step of applying, at the at least one encoder processor, a hierarchical vector quantization compression algorithm to the digitized image signal. At the next step, a resultant encoded bit stream generated by the application of the algorithm is collected. The method and apparatus of the present invention may be used in conjunction with an ordinary modem to transmit and/or receive audio, video sequences or still images.

Abstract: A method and apparatus for encoding a sampled signal using a nonscalar vector quantizer. A lapping window operator is applied to blocks of signal samples of length greater than N to produce length-N transform vectors which are then encoded by a nonscalar vector quantizer. The method and apparatus may use or have a window operator that is identical to that of the modulated lapped transform, that is a 2N.times.N matrix operator, that is identical to that of the modulated lapped transform with asymmetrical, orthogonal butterflies, or that has butterfly transmittances given by h?n!=sin?(2n+1).pi./4N!. The method and apparatus may use or have a vector quantizer is an unconstrained mean-square optimized vector quantizer having codewords of length N.

Abstract: A method and apparatus for recognition of objects such as faces in images using signal compression techniques (i.e., coding techniques) in which a portion of the image which includes the object to be recognized (e.g., the face) is coded, and the resultant coded data is matched against previously coded and stored training data which makes up a known object database. A given object in an input image signal is matched to one of a plurality of known objects stored in a database, wherein the stored representation of each of the known objects comprises a codebook generated based on training image signals comprising the known object. A first illustrative embodiment comprises the steps of decomposing the given object into blocks; performing a plurality of encodings of the given object, each encoding comprising coding the object with use of one of the codebooks; determining a coding error for each encoding; and matching the given object to one of the known objects based on the coding errors.

Abstract: An N-stage vector quantizer in which increasingly smaller portions of a vector are compared to a threshold until the quantization error is less than the threshold. The threshold may be adaptive to insure a constant bit rate. The first stage performs VQ and inverse VQ on an input vector. The difference between the inverse VQ and the input vector is determined to create a first stage residual error. If the first stage residual error is less than a threshold, no further stages of the multistage vector quantizer are used and the input vector or first stage residual is passed on to the output stage. However, if the first stage residual error is not less than the threshold, the residual error is passed to stage two of the multistage vector quantizer where VQ and inverse VQ are performed on the first stage residual error. The process is continued until the residual is less than the threshold.

Abstract: An image information coding apparatus includes a cut-out or extracting unit for cutting out image information into blocks each having a predetermined size, e.g., 4.times.4 pixels, a preprocessing unit such as an orthogonal transformation unit (e.g., a Hadamard transformation unit) and a scalar quantization unit for performing predetermined preprocessing of the block as a matrix, a separator for separating a plurality of components (e.g., phase information, amplitude information, and zero information) from each element of the preprocessed matrix, information quantizers for coding individual pieces of information constituted by each component, and a synthesizer for synthesizing coding results.

Abstract: The quality of reconstructed images is improved by a post-processing method at the decoder which uses codeword-dependent, spatially-varying 2-dimensional filters. The post-filtering, working across block boundaries, reduces coding artifacts and increases the signal-to-noise ratio. The technique is applicable for use with any VQ-based image compression and can also be used in conjunction with image magnification.

Abstract: An image compression coding method of a digital image transmission system utilizes a known vector quantization process.

Abstract: A method of condensing color image data in an image processing system so as to simplify operations while reducing errors in the condensation.The method includes, by computer, generating electrical signals corresponding to color image data representative of a physical image; associating an electrical signal corresponding to a representative code with portions of the signals representative of data in the color image data corresponding to each of the pixels; grouping together electrical signals representative of pixels having identical representative codes to form at least one group; assigning to the representative code for each of the at least one group a condensed pixel representable by the representative code for that group; and setting an electrical signal portion corresponding to each pixel in that group equal to an electrical signal representative of a pixel corresponding to the condensed pixel.