Showing papers on "Vector quantization published in 1991"
Book•
1 Jan 1991TL;DR: The author explains the design and implementation of the Levinson-Durbin Algorithm, which automates the very labor-intensive and therefore time-heavy and expensive process of designing and implementing a Quantizer.
Abstract: 1 Introduction- 11 Signals, Coding, and Compression- 12 Optimality- 13 How to Use this Book- 14 Related Reading- I Basic Tools- 2 Random Processes and Linear Systems- 21 Introduction- 22 Probability- 23 Random Variables and Vectors- 24 Random Processes- 25 Expectation- 26 Linear Systems- 27 Stationary and Ergodic Properties- 28 Useful Processes- 29 Problems- 3 Sampling- 31 Introduction- 32 Periodic Sampling- 33 Noise in Sampling- 34 Practical Sampling Schemes- 35 Sampling Jitter- 36 Multidimensional Sampling- 37 Problems- 4 Linear Prediction- 41 Introduction- 42 Elementary Estimation Theory- 43 Finite-Memory Linear Prediction- 44 Forward and Backward Prediction- 45 The Levinson-Durbin Algorithm- 46 Linear Predictor Design from Empirical Data- 47 Minimum Delay Property- 48 Predictability and Determinism- 49 Infinite Memory Linear Prediction- 410 Simulation of Random Processes- 411 Problems- II Scalar Coding- 5 Scalar Quantization I- 51 Introduction- 52 Structure of a Quantizer- 53 Measuring Quantizer Performance- 54 The Uniform Quantizer- 55 Nonuniform Quantization and Companding- 56 High Resolution: General Case- 57 Problems- 6 Scalar Quantization II- 61 Introduction- 62 Conditions for Optimality- 63 High Resolution Optimal Companding- 64 Quantizer Design Algorithms- 65 Implementation- 66 Problems- 7 Predictive Quantization- 71 Introduction- 72 Difference Quantization- 73 Closed-Loop Predictive Quantization- 74 Delta Modulation- 75 Problems- 8 Bit Allocation and Transform Coding- 81 Introduction- 82 The Problem of Bit Allocation- 83 Optimal Bit Allocation Results- 84 Integer Constrained Allocation Techniques- 85 Transform Coding- 86 Karhunen-Loeve Transform- 87 Performance Gain of Transform Coding- 88 Other Transforms- 89 Sub-band Coding- 810 Problems- 9 Entropy Coding- 91 Introduction- 92 Variable-Length Scalar Noiseless Coding- 93 Prefix Codes- 94 Huffman Coding- 95 Vector Entropy Coding- 96 Arithmetic Coding- 97 Universal and Adaptive Entropy Coding- 98 Ziv-Lempel Coding- 99 Quantization and Entropy Coding- 910 Problems- III Vector Coding- 10 Vector Quantization I- 101 Introduction- 102 Structural Properties and Characterization- 103 Measuring Vector Quantizer Performance- 104 Nearest Neighbor Quantizers- 105 Lattice Vector Quantizers- 106 High Resolution Distortion Approximations- 107 Problems- 11 Vector Quantization II- 111 Introduction- 112 Optimality Conditions for VQ- 113 Vector Quantizer Design- 114 Design Examples- 115 Problems- 12 Constrained Vector Quantization- 121 Introduction- 122 Complexity and Storage Limitations- 123 Structurally Constrained VQ- 124 Tree-Structured VQ- 125 Classified VQ- 126 Transform VQ- 127 Product Code Techniques- 128 Partitioned VQ- 129 Mean-Removed VQ- 1210 Shape-Gain VQ- 1211 Multistage VQ- 1212 Constrained Storage VQ- 1213 Hierarchical and Multiresolution VQ- 1214 Nonlinear Interpolative VQ- 1215 Lattice Codebook VQ- 1216 Fast Nearest Neighbor Encoding- 1217 Problems- 13 Predictive Vector Quantization- 131 Introduction- 132 Predictive Vector Quantization- 133 Vector Linear Prediction- 134 Predictor Design from Empirical Data- 135 Nonlinear Vector Prediction- 136 Design Examples- 137 Problems- 14 Finite-State Vector Quantization- 141 Recursive Vector Quantizers- 142 Finite-State Vector Quantizers- 143 Labeled-States and Labeled-Transitions- 144 Encoder/Decoder Design- 145 Next-State Function Design- 146 Design Examples- 147 Problems- 15 Tree and Trellis Encoding- 151 Delayed Decision Encoder- 152 Tree and Trellis Coding- 153 Decoder Design- 154 Predictive Trellis Encoders- 155 Other Design Techniques- 156 Problems- 16 Adaptive Vector Quantization- 161 Introduction- 162 Mean Adaptation- 163 Gain-Adaptive Vector Quantization- 164 Switched Codebook Adaptation- 165 Adaptive Bit Allocation- 166 Address VQ- 167 Progressive Code Vector Updating- 168 Adaptive Codebook Generation- 169 Vector Excitation Coding- 1610 Problems- 17 Variable Rate Vector Quantization- 171 Variable Rate Coding- 172 Variable Dimension VQ- 173 Alternative Approaches to Variable Rate VQ- 174 Pruned Tree-Structured VQ- 175 The Generalized BFOS Algorithm- 176 Pruned Tree-Structured VQ- 177 Entropy Coded VQ- 178 Greedy Tree Growing- 179 Design Examples- 1710 Bit Allocation Revisited- 1711 Design Algorithms- 1712 Problems
8,059 citations
TL;DR: It is demonstrated that for very noisy channels and a heavily correlated source, when the code book size is large, the number of encoding regions is considerably smaller than the codebook size-implying a reduction in encoding complexity.
Abstract: The performance and complexity of channel-optimized vector quantizers are studied for the Gauss-Markov source. Observations on the geometric structure of these quantizers are made, which have an important implication on the encoding complexity. For the squared-error distortion measure, it is shown that an operation equivalent to a Euclidean distance measurement with respect to an appropriately defined set of points (used to identify the encoding regions) can be used to perform the encoding. This implies that the encoding complexity is proportional to the number of encoding regions. It is then demonstrated that for very noisy channels and a heavily correlated source, when the codebook size is large, the number of encoding regions is considerably smaller than the codebook size-implying a reduction in encoding complexity. >
314 citations
14 Apr 1991
TL;DR: It is shown that the split vector quantizer can quantize LPC information in 24 b/frame with 1-dB average spectral distortion and <2% outlier frames (having spectral distortion greater than 2 dB).
Abstract: Though vector quantizers are more efficient than scalar quantizers, their use for fine quantization of linear predictive coding (LPC) information (using 24-26 b/frame) is impeded due to their prohibitively high complexity. In the present work, a split vector quantization approach is used to overcome the complexity problem. The LPC vector, consisting of ten line spectral frequencies (LSFs), is divided into two parts and each part is quantized separately using vector quantization. Using the localized spectral sensitivity property of the LSF parameters, a weighted LSF distance measure is proposed. Using this distance measure, it is shown that the split vector quantizer can quantize LPC information in 24 b/frame with 1-dB average spectral distortion and >
167 citations
Patent•
10 Jun 1991
TL;DR: In this paper, an image data compression technique is described which utilizes calculating means and a selected series of bit calculating stages having delays, to estimate one or more quantization parameters for such data.
Abstract: An image data compression technique is described which utilizes calculating means and a selected series of bit calculating stages having delays, to estimate one or more quantization parameters for such data. The estimation process preferably is iterated a number of times, with the values found through each estimation being used as the trial values for subsequent estimations. In addition, an initial trial value is selected by a data look ahead technique, which assures that its value is within range of the final quantization parameter used to quantize the data. The final quantization parameter insures that the compressed data fits within a predetermined number of encoded data bits to be transmitted or recorded, for example, in a recording medium.
146 citations
TL;DR: The results show that even with a short sequence of only four isolated digits, a speaker can be verified with an average equal-error rate of less than 3 %, and the small improvement over the vector quantization approach indicates the weakness of the Markovian transition probabilities for characterizing speaker-dependent transitional information.
Abstract: Linear predictive hidden Markov models have proved to be efficient for statistically modeling speech signals. The possible application of such models to statistical characterization of the speaker himself is described and evaluated. The results show that even with a short sequence of only four isolated digits, a speaker can be verified with an average equal-error rate of less than 3 %. These results are slightly better than the results obtained using speaker-dependent vector quantizers, with comparable numbers of spectral vectors. The small improvement over the vector quantization approach indicates the weakness of the Markovian transition probabilities for characterizing speaker-dependent transitional information. >
132 citations
1 Jan 1991
TL;DR: This chapter discusses efficient statistical computations for optimal color quantization based on variance minimization, a 3D clustering process that leads to significant image data compression, making extra frame buffer available for animation and reducing bandwidth requirements.
Abstract: Publisher Summary This chapter discusses efficient statistical computations for optimal color quantization. Color quantization is a must when using an inexpensive 8-bit color display to display high-quality color images. Even when 24-bit full color displays become commonplace in the future, quantization will still be important because it leads to significant image data compression, making extra frame buffer available for animation and reducing bandwidth requirements. Color quantization is a 3D clustering process. A color image in an RGB mode corresponds to a three-dimensional discrete density. In this chapter, quantization based on variance minimization is discussed. Efficient computations of color statistics are described. An optimal color quantization algorithm is presented. The algorithm was implemented on a SUN 3/80 workstation. It took only 10 s to quantize a 256 × 256 image. The impact of optimizing partitions is very positive. The new algorithm achieved, on average, one-third and one-ninth of mean-square errors for the median-cut and Wan et. al. algorithms, respectively.
121 citations
14 Apr 1991
TL;DR: A fast nearest-neighbor search algorithm is developed which incorporates prior information about input vectors in the form of a vector from the codebook which is known to be near the input vector, though it may not be the nearest codebook vector.
Abstract: A fast nearest-neighbor search algorithm is developed which incorporates prior information about input vectors. The prior information comes in the form of a vector from the codebook which is known to be near the input vector, though it may not be the nearest codebook vector. A number of applications are described for which such prior information is available. The algorithm has a very simple structure and can be designed to have very low memory requirements. The new algorithm requires much less computation for constructing precomputed tables than previously proposed algorithms with comparable performance. Simulations show dramatic saving over conventional full search methods. >
120 citations
TL;DR: Trellis-coded quantization is generalized to allow a vector reproduction alphabet and it is shown that for a stationary ergodic vector source, the quantization noise is zero-mean and of a variance equal to the difference between the source variance and the variance of the reproduction sequence.
Abstract: Trellis-coded quantization is generalized to allow a vector reproduction alphabet. Three encoding structures are described, several encoder design rules are presented, and two design algorithms are developed. It is shown that for a stationary ergodic vector source, if the optimized trellis-coded vector quantization reproduction process is jointly stationary and ergodic with the source, then the quantization noise is zero-mean and of a variance equal to the difference between the source variance and the variance of the reproduction sequence. Several examples illustrate the encoder design procedure and performance. >
116 citations
Proceedings Article•
[...]
TL;DR: The author surveys the menagerie of quantization and compression algorithms in the specific context of image compression and provides some general comparisons based on performance, complexity, and side benefits of particular coding techniques.
Abstract: Summary form only given. The author surveys the menagerie of quantization and compression algorithms in the specific context of image compression and provides some general comparisons based on performance, complexity, and side benefits of particular coding techniques. >
106 citations
TL;DR: A stochastic Lyapunov argument shows that competitive synaptic vectors converge to centroids exponentially quickly and reduces competitive learning to stochastically gradient descent, extending to competitive estimation of local covariances and higher order statistics.
Abstract: Competitive learning systems are examined as stochastic dynamical systems. This includes continuous and discrete formulations of unsupervised, supervised, and differential competitive learning systems. These systems estimate an unknown probability density function from random pattern samples and behave as adaptive vector quantizers. Synaptic vectors, in feedforward competitive neural networks, quantize the pattern space and converge to pattern class centroids or local probability maxima. A stochastic Lyapunov argument shows that competitive synaptic vectors converge to centroids exponentially quickly and reduces competitive learning to stochastic gradient descent. Convergence does not depend on a specific dynamical model of how neuronal activations change. These results extend to competitive estimation of local covariances and higher order statistics. >
98 citations
14 Apr 1991
TL;DR: Two language recognition algorithms are proposed and some experimental results are described, based on a single universal (common) VQ codebook for all languages, and its occurrence probability histograms.
Abstract: Two language recognition algorithms are proposed and some experimental results are described. While many studies have been done concerning the speech recognition problem, few studies have addressed the language recognition task. The speech data used contains 20 languages: 16 sentences uttered twice by 4 males and 4 females. The duration of each sentence is about 8 seconds. The first algorithm is based on the standard vector quantization (VQ) technique. Every language is characterized by its own VQ codebook. The second algorithm is based on a single universal (common) VQ codebook for all languages, and its occurrence probability histograms. Every language is characterized by a histogram. The experiment results show that the recognition rates for the first and second algorithms were 65% and 80%, respectively, each using just 8 sentences of unknown speech (about 64 seconds). >
TL;DR: In this article, a classification algorithm in the discrete cosine transform (DCT) domain for the classified vector quantization (CVQ) technique is proposed, which employs four DCT coefficients of 4*4 subblock as edge-oriented features.
Abstract: A classification algorithm in the discrete cosine transform (DCT) domain for the classified vector quantization (CVQ) technique is proposed. The classifier employs four DCT coefficients of 4*4 subblock as edge-oriented features. The classifier is designed using a cluster-seeking algorithm to ensure that the centroid of a set of vectors in a class always belong to that class. Since the classification is performed in the DCT domain, this approach can be easily extended to the DCT transform coding technique. Simulation results show that a good visual quality of the coded image at fixed rates in the 0.625-0.813 b/pixel (bpp) range is obtained with comparable complexity. The weighted MSE (WMSE) analysis in conjunction with the proposed classifier is discussed. >
TL;DR: A system for the compression of HDTV for DVTR's based on this rate-constrained optimal block-adaptive technique is designed using the DCT and vector quantization with a multistage structure.
Abstract: An image coding algorithm for digital video tape recorders (DVTR) must satisfy several requirements which do not arise in other applications of video compression. A key constraint on the data format is satisfied if every frame (or field) of video is partitioned into a small number of subimages, each independently coded with a fixed number of bits. This requirement excludes the use of interframe coding and most variable-rate image coding algorithms. We propose to use a new algorithm that codes a subimage efficiently under this data format constraint and allows virtually lossless reproduction at reasonable low bit rates. Each subimage is partitioned into non-overlapping blocks and each block is coded by one of a finite set of predesigned block quantizers covering a range of bit rates. For the ith block in the subimage, a rate function Ri(Li) and a distortion function Di(Li) is tabulated for each block quantizer Li. A near-optimal quantizer allocation algorithm based on the Lagrange Multiplier method is used to select a particular quantizer for each block. The objective is to minimize the distortion of the entire subimage under the constraint on the total number of bits for the subimage. A system for the compression of HDTV for DVTR's based on this rate-constrained optimal block-adaptive technique is designed using the DCT and vector quantization with a multistage structure. Simulation results demonstrate that this algorithm has the potential of achieving virtually lossless compression for digital tape recording of HDTV.
TL;DR: Simulations showed that unsupervised DCL- trained synaptic vectors converged to class centroids at least as fast as, and wandered less about, SCL-trained synaptic vectors did, which favored DCL over SCL for classification accuracy.
Abstract: A comparison is made of a differential-competitive-learning (DCL) system with two supervised competitive-learning (SCL) systems for centroid estimation and for phoneme recognition. DCL provides a form of unsupervised adaptive vector quantization. Standard stochastic competitive-learning systems learn only if neurons win a competition for activation induced by randomly sampled patterns. DCL systems learn only if the competing neurons change their competitive signal. Signal-velocity information provides unsupervised local reinforcement during learning. The sign of the neuronal signal derivative rewards winners and punishes losers. Standard competitive learning ignores instantaneous win-rate information. Synaptic fan-in vectors adaptively quantize the randomly sampled pattern space into nearest-neighbor decision classes. More generally, the synaptic-vector distribution estimates the unknown sampled probability density function p(x). Simulations showed that unsupervised DCL-trained synaptic vectors converged to class centroids at least as fast as, and wandered less about these centroids than, SCL-trained synaptic vectors did. Simulations on a small set of English phonemes favored DCL over SCL for classification accuracy. >
8 Apr 1991
TL;DR: A new vector quantizer permits logarithmic-time encoding and yet performs better than the locally optimal quantizers generated by the LBG algorithm.
Abstract: A new vector quantizer permits logarithmic-time encoding and yet performs better than the locally optimal quantizers generated by the LBG algorithm. The success is credited to an elaborated tree-structured optimization process in the codebook design. >
Patent•
AT&T1
TL;DR: In this paper, a geometric vector quantizer coding technique is illustrated in the context of a full motion video coder based on a three-dimensional sub-band framework, which is decomposed into different spatial-temporal frequency bands and based on the data in each band, different quantization strategies are applied to the bands.
Abstract: A geometric vector quantizer coding technique is illustrated in the context of a full motion video coder based on a three-dimensional sub-band framework. The original image data is decomposed into different spatial-temporal frequency bands and based on the data in each band, different quantization strategies are applied to the bands. The lowest spatial-temporal frequency band is coded using a traditional three-dimensional switched predictor and optimum scaler quantizer. The non-dominant sub-bands are coded using the vector quantization approach to efficiently encode the images while appropriately exploiting the sparse, highly structured nature of the data to design the codebooks. Variable resolution is achieved using these techniques and no training is needed to establish or maintain the codebook. A fast codebook search technique is provided for identifying the optimal codebook vector for representing a block of input data. Examples of two and three level geometric vector quantizers are also provided.
Patent•
IBM1
TL;DR: In this paper, a scaling factor for the quantization tables of the multiple image components is defined, and the scaling factor signals changes in quantization for successive blocks of the image data.
Abstract: A system and method for masking adaptive quantization during compressed image data transmission by defining a scaling factor for the quantization tables of the multiple image components, wherein the scaling factor signals changes in quantization for successive blocks of the image data. The scaling factor is transmitted as a further component together with the image components to thereby signal adaptive quantization of the image data.
14 Apr 1991
TL;DR: The purpose of this work is to propose a new scheme for vectors quantization of wavelet coefficients based on lattice vector quantization, and the application of the D/sub 4/, E/sub 8/ and Barnes-Wall Lambda /sub 16/ lattices is investigated.
Abstract: An image coding scheme has been introduced by the authors (see IEEE ICASSP, p.2297, 1990). This scheme involves two steps. A biorthogonal wavelet transform is applied to the original image, and wavelet coefficients are then vector quantized using the LBG (Linde, Buzo and Gray, 1980) method. The purpose of this work is to propose a new scheme for vector quantization of wavelet coefficients. The proposed method is based on lattice vector quantization. The application of the D/sub 4/, E/sub 8/ and Barnes-Wall Lambda /sub 16/ lattices is investigated. These lattices are used to encode wavelet coefficients whose PDFs are close to Laplacian. A variable-length coding method is applied and the trade-off between distortion and optimal rate is investigated. Experimental results on the Lena image using the Lambda /sub 16/ lattice leads to a peak signal-to-noise ratio (PSNR) of 31.14 dB at 0.08 bpp. This result outperforms, to the authors knowledge, all other methods. Edges which are most of interest for image analysis are particularly sharp without any smoothing artefacts. >
1 Jun 1991
TL;DR: In this article, the vector quantization algorithm proposed by Equitz was applied to the problem of efficiently selecting colors for a limited image palette, and the algorithm performed the quantization by merging pairwise nearest neighbor (PNN) clusters.
Abstract: We apply the vector quantization algorithm proposed by Equitz to the problem of efficiently selecting colors for a limited image palette. The algorithm performs the quantization by merging pairwise nearest neighbor (PNN) clusters. Computational efficiency is achieved by using k- dimensional trees to perform fast PNN searches. In order to reduce the number of initial image colors, we first pass the image through a variable-size cubical quantizer. The centroids of colors that fall in each cell are then used as sample vectors for the merging algorithm. Tremendous computational savings is achieved from this initial step with very little loss in visual quality. To account for the high sensitivity of the human visual system to quantization errors in smoothly varying regions of an image, we incorporate activity measures both at the initial quantization step and at the merging step so that quantization is fine in smooth regions and coarse in active regions. The resulting images are of high visual quality. The computation times are substantially smaller than that of the iterative Lloyd-Max algorithm and are comparable to a binary splitting algorithm recently proposed by Bouman and Orchard.
11 Jun 1991
TL;DR: The authors summarize a speaker adaptation algorithm based on codebook mapping from one speaker to a standard speaker to be useful in various kinds of speech recognition systems such as hidden-Markov-model-based, feature- based, and neural-network-based systems.
Abstract: The authors summarize a speaker adaptation algorithm based on codebook mapping from one speaker to a standard speaker. This algorithm has been developed to be useful in various kinds of speech recognition systems such as hidden-Markov-model-based, feature-based, and neural-network-based systems. The codebook mapping speaker adaptation algorithm has been much improved by introducing several ideas based on fuzzy vector quantization. This fuzzy codebook mapping algorithm is also applicable to voice conversion between arbitrary speakers. >
Patent•
4 Nov 1991TL;DR: In this article, a process for adaptive quantization with quantization errors which have little disturbing visual effects, using a digital block-related process for data reduction in digital images or image sequences, is described.
Abstract: A process for adaptive quantization with quantization errors which have little disturbing visual effects, using a digital block-related process for data reduction in digital images or image sequences. It is provided in this process that an image to be transmitted is subdivided into a multiplicity of blocks, and that there is calculated for each block a parameter for setting the quantization assigned to this block. The calculation of this parameter is performed in this case with the aid of a subdivision of each block into subregions, there being calculated for each subregion an activity measure with the aid of which a quantization parameter is determined for each subregion. Finally, the quantization parameters of all subregions are summed up within the blocks in the case of all blocks and multiplicatively scaled.
TL;DR: An architecture suitable for real-time image coding using adaptive vector quantization (VQ) is presented, where the data is accessed simultaneously and in parallel on the basis of its content.
Abstract: An architecture suitable for real-time image coding using adaptive vector quantization (VQ) is presented. This architecture is based on the concept of content-addressable memory (CAM), where the data is accessed simultaneously and in parallel on the basis of its content. VQ essentially involves, for each input vector, a search operation to obtain the best match codeword. A speedup results if a CAM-based implementation is used. This speedup, coupled with the gains in execution time for the basic distortion operation, implies that even codebook generation is possible in real time ( >
TL;DR: A hybrid BTC-VQ-DCT (block truncation coding, vector quantization, and discrete cosine transform) image coding algorithm is presented that combines the simple computation and edge preservation properties of BTC with the high fidelity and high-compression ratio of adaptive DCT and good subjective performance of VQ.
Abstract: A hybrid BTC-VQ-DCT (block truncation coding, vector quantization, and discrete cosine transform) image coding algorithm is presented. The algorithm combines the simple computation and edge preservation properties of BTC and the high fidelity and high-compression ratio of adaptive DCT with the high-compression ratio and good subjective performance of VQ, and can be implemented with significantly lower coding delays than either VQ or DCT alone. The bit-map generated by BTC is decomposed into a set of vectors which are vector quantized. Since the space of the BTC bit-map is much smaller than that of the original 8-b image, a lookup-table-based VQ encoder has been designed to 'fast encode' the bit-map. Adaptive DCT coding using residual error feedback is implemented to encode the high-mean and low-mean subimages. The overall computational complexity of BTC-VQ-DCT coding is much less than either DCT and VQ, while the fidelity performance is competitive. The algorithm has strong edge-preserving ability because of the implementation of BTC as a precompress decimation. The total compression ratio is about 10:1. >
Patent•
19 Nov 1991
TL;DR: In this paper, a variable rate vector quantization method employs a tree structured codebook and the level of the codebook from which codevectors are selected is determined by a threshold.
Abstract: A variable rate vector quantization method employs a tree structured codebook. The level of the codebook from which codevectors are selected is determined by a threshold. The threshold varies according to the fullness of a buffer which stores vector quantized data to be transmitted.
NEC1
TL;DR: In this paper, vector quantization is carried out through a plurality of vector quantizers which are connected in cascade to one another through subtractors and which cooperate with code books, respectively.
Abstract: In a speech parameter encoding method of encoding an input speech signal into a sequence of encoded signals, vector quantization is carried out through a plurality of vector quantizers which are connected in cascade to one another through subtractors and which cooperate with code books, respectively. Such a cascade connection of the vector quantizers may be supplied with the input speech signal either at every frame or at every subframe shorter than the frame to produce a set of code vector candidates from each of the vector quantizers. Each set of the code vector candidates is sent to a cumulative distortion calculator to select an optimum combination of the code vector candidates and to produce, as the encoded signals, an index representative of the optimum combination. Alternatively, a sequence of coefficients which represent a spectrum parameter is calculated at every frame and is divided into a plurality of coefficient groups each of which is subjected to vector quantization to produce a plurality of code vector candidates for each coefficient group. Cumulative distortions are calculated for combinations of the code vector candidates to select an optimum combination.
TL;DR: A codebook sharing technique, called constrained storage vector quantization (CSVQ), is introduced, which offers a convenient and optimal way of trading off performance against storage.
Abstract: A codebook sharing technique, called constrained storage vector quantization (CSVQ), is introduced. This technique offers a convenient and optimal way of trading off performance against storage. The technique can be used in conjunction with tree-structured vector quantization (VQ) and other structured VQ techniques that alleviate the search complexity obstacle. The effectiveness of CSVQ is illustrated for coding transform coefficients of audio signals with multistage VQ. >
Patent•
19 Nov 1991TL;DR: In this article, a variable rate vector quantization apparatus and method employs a tree structured codebook, where code vectors are selected from different levels of the codebook according to the value of a threshold.
Abstract: A variable rate vector quantization apparatus and method employs a tree structured codebook. Code vectors are selected from different levels of the codebook according to the value of a threshold. The value of the threshold is periodically adjusted according to the fullness of a buffer that stores vector quantized data to be transmitted. According to the invention, vector quantized data for redundant, or similar, vectors is not transmitted. Rather, a "copy last vector" instruction is transmitted for these vectors to achieve further data compression. A method of mean removal from vectors to be vector quantized is also disclosed.
Patent•
27 Sep 1991TL;DR: In this article, an image signal coding/decoding system consisting of an encoder for encoding and decoding the coded image signal by using the adaptive inverse-quantization is presented, where the encoder comprises an activity detection element for detecting, every blocks obtained by dividing the image signal every a specific time length, activities indicating degree of changes in the signal in the respective blocks.
Abstract: An image signal coding/decoding system comprises an encoder for encoding an image signal by using the adaptive quantization, and a decoder for decoding the coded image signal by using the adaptive inverse-quantization The encoder comprises an activity detection element for detecting, every blocks obtained by dividing an image signal every a specific time length, activities indicating degree of changes in the signal in the respective blocks; a quantization class determination element for determining class values of a plurality of quantization classes set in correspondence with respective stages of the adaptive quantization from the detected activities; and for outputting the determined class values as class information to the decoder, a filter element for filtering not only the class values but also class values at the peripheral blocks; and a quantization step width determination element for determining, every blocks, widths of quantization steps corresponding to a value obtained by multiplying these class values by fixed coefficients on the basis of class values subjected to filtering The decoder comprises a decode element for decoding coded data transmitted from the encoder by using the variable-length codes, and an inverse-quantization element for inverse-quantizing the variable-length decoded signal of the basis of class information transmitted from the quantization class determination element
TL;DR: A speech analysis-synthesis system having a spectral envelope generator for generating a spectral enclosure which is so smooth that excessive beating is avoided, a spectral envelopes vector converter for sampling the spectral envelope at equal intervals on mel-scale, and a vector quantizer for quantizing vectors are provided.
Abstract: A vector quantizing apparatus having a general vector quantization circuit, and a storage means for storing at least one frame of data as the result of comparison by a matching circuit is provided. Further, provided are a speech analysis-synthesis system having a spectral envelope generator for generating a spectral envelope which is so smooth that excessive beating is avoided, a spectral envelope vector converter for sampling the spectral envelope at equal intervals on mel-scale, a vector quantizer for quantizing vectors, and a spectral envelope reconstructor for reconstructing the spectral envelope by interpolation on the basis of combined parabolas.
TL;DR: It is found that when rate, distortion, arithmetic complexity, and storage are all taken into account, FCVQ outperforms TSVQ in a number of cases, at the expense of a slight increase in distortion and a substantial increase in storage.
Abstract: A fast method for searching an unstructured vector quantization (VQ) codebook is introduced and analyzed. The method, fine-coarse vector quantization (FCVQ), operates in two stages: a 'fine' structured VQ followed by a table lookup 'coarse' unstructured VQ. Its rate, distortion, arithmetic complexity, and storage are investigated using analytical and experimental means. Optimality condition and an optimizing algorithm are presented. The results of experiments with both uniform scalar quantization and tree-structured VQ (TSVQ) as the first stage are reported. Comparisons are made with other fast approaches to vector quantization, especially TSVQ. It is found that when rate, distortion, arithmetic complexity, and storage are all taken into account, FCVQ outperforms TSVQ in a number of cases. In comparison to full search quantization, FCVQ has much lower arithmetic complexity, at the expense of a slight increase in distortion and a substantial increase in storage. The increase in mean-squared error (over full search) decays as a negative power of the available storage. >