Vector Quantization Codebook Research Articles

A general codebook design method for vector quantization

Vector quantization (VQ) is widely used in image processing applications, the primary focus of VQ is to determine a codebook to represent the original image well. In order to make a codebook perform better on both distortion and bit rate (BR), a general codebook (GCB) for VQ is proposed in this paper. Unlike common codebook (CCB) or private codebook (PCB), GCB is a new structure of codebook where the codewords can either come from CCB or by training the input image. By applying the codewords in CCB that perform well and updating inactive codewords, only the new generated codewords and flags of codewords to be replaced are transmitted along with index table (IT). Therefore,the BR can be significantly reduced while the performance of distortion can be efficiently improved. The experimental results demonstrate that our proposed GCB has a better performance than CCB and various kinds of PCB-based methods.

Reversible data hiding for VQ indices using hierarchical state codebook mapping

In this paper, we present an approach to efficiently hide sensitive data in vector quantization (VQ) indices and reversibly extract sensitive data from encrypted code stream. The approach uses two patterns to compress VQ indices. When an index is equal to its upper neighbor’s index or left neighbor’s index, it is encoded by the corresponding equivalent index; otherwise, it is encoded by a modified VQ codebook mapping named as hierarchical state codebook mapping (HSCM). In the proposed scheme, the hierarchical state codebook mapping (HSCM) is main coding pattern and it is generated according to the side-match distortion method(SMD). By the above two patterns, the size of original code stream is reduced, and more storage space can be used to embed sensitive data. The experimental results indicated that the proposed scheme can achieve a higher embedding capacity than the previous state-of-the-art VQ-index-based data hiding methods.

A Novel and Efficient Vector Quantization Based CPRI Compression Algorithm

The future wireless network, such as the Centralized Radio Access Network (C-RAN), will need to deliver data rate about 100–1000 times the current fourth-generation (4G) technology. For the C-RAN-based network architecture, there is a pressing need for tremendous enhancement of the effective data rate of the common public radio interface (CPRI). Compression of CPRI data is one of the potential enhancements. In this paper, we introduce a vector quantization based compression algorithm for CPRI links, utilizing the Lloyd algorithm. Methods to vectorize the I/Q samples and enhanced initialization of the Lloyd algorithm for codebook training are investigated for improved performance. Multistage vector quantization and unequally protected multigroup quantization are considered to reduce codebook search complexity and codebook size. Simulation results show that our solution can achieve compression of four times for uplink and 4.5 times for downlink, within $2\%$ error vector magnitude (EVM) distortion. Remarkably, vector quantization codebook proves to be quite robust against data modulation mismatch, fading, signal-to-noise ratio (SNR), and Doppler spread.

VQ codebook design using modified K-means algorithm with feature classification and grouping based initialization

Vector quantization (VQ) has been successfully used in data compression and feature extraction areas. Codebook design is the essential step of VQ. The K-means algorithm is a famous data clustering technique which is also an efficient codebook design scheme. The main disadvantages of K-means algorithm lie in that the initial cluster centroids greatly affect the convergence speed and the final clustering performance. In the past two decades, many codebook initialization techniques have been proposed. However, most of these techniques do not make full use of the features of the training vectors, and some techniques require high extra computational load. This paper presents an efficient and simple technique for the conventional K-means algorithm based on feature classification and grouping. Firstly, all training vectors are classified into sixteen categories based on a two-level classifier including an edge classifier and a contrast classifier. Then, the training vectors in each category are sorted based on their norm values and divided into groups. Each group has the same size, and the centroid vector of each group is calculated as an initial codeword. Experimental results show that, compared with several typical initialization techniques, our technique can obtain a better codebook along with a faster convergence speed in a shorter time.

Rapid Generation of the State Codebook in Side Match Vector Quantization

Rapid Generation of the State Codebook in Side Match Vector Quantization

Quantization and Feedback of Principal Modes for Dispersion Mitigation and Multiplexing in Multimode Fibers

The larger core diameter of multimode fibers (MMFs) allows several propagating modes, but differential group delays and modal cross talk cause modal dispersion, thus limiting the data rate. Recent results have shown the existence of a set of principal modes (PMs) that allow dispersion free communication to the first order. Moreover, the linear independence of the PMs also permits multiplexing in MMFs. To effectively utilize these PMs at the transmitter, it is essential to estimate them and feed them back to the transmitter. In this paper, we propose quantization schemes that enable efficient encapsulation of the PMs for feedback for MMFs with a small number of modes. For shorter links, we use vector quantization codebooks based on the Linde-Buzo–Gray (LBG) algorithm, while for long links with low mode dependent losses (MDL), we show analytically that the PMs are Haar distributed, and use Grassmannian line packing based codebooks. Simulations reveal that the quantization limits dispersion and the performance of the quantized PMs is within 2 dB of the ideal principal modes with only 6 bits quantization, both with and without MDL. In addition, the quantized PMs permit efficient multiplexing with low crosstalk in both the lossless and MDL cases.

Limited-Feedback Massive MISO Systems With Trellis-Coded Quantization for Correlated Channels

In this paper, we propose trellis coded quantization (TCQ) based limited feedback techniques for massive multiple-input single-output (MISO) frequency division duplexing (FDD) systems in temporally and spatially correlated channels. We exploit the correlation present in the channel to effectively quantize channel direction information (CDI). For multiuser (MU) systems with matched-filter (MF) precoding, we show that the number of feedback bits required by the random vector quantization (RVQ) codebook to match even a small fraction of the perfect CDI signal-to-interference-plus-noise ratio (SINR) performance is large. With such large numbers of bits, the exhaustive search required by conventional codebook approaches make them infeasible for massive MISO systems. Motivated by this, we propose a differential TCQ scheme for temporally correlated channels that transforms the source constellation at each stage in a trellis using 2D translation and scaling techniques. We derive a scaling parameter for the source constellation as a function of the temporal correlation and the number of BS antennas. We also propose a TCQ based limited feedback scheme for spatially correlated channels where the channel is quantized directly without performing decorrelation at the receiver. Simulation results show that the proposed TCQ schemes outperform the existing noncoherent TCQ (NTCQ) schemes, by improving the spectral efficiency and beamforming gain of the system. The proposed differential TCQ also reduces the feedback overhead of the system compared to the differential NTCQ method.

An Adaptive Orthogonal M-Split Initialization Method for VQ Codebook Generation

Linde–Buzo–Gray (LBG) algorithm is a universal method to design codebook in vector quantization(VQ). This paper proposed an adaptive orthogonal M-split initialization method to improve the computational efficiency of LBG algorithm. The method splits one code word into 2, 4 or 5 new code words with adaptive split coefficient vectors and set the increment to be orthogonal in 4-split and 5-split situations, aiming at decreasing the iterations of the following clustering. Experiment is conducted on both TIMIT and RASC863 speech database, which shows that the proposed algorithm provides a reduction of 18%~45% in designing codebook in size of 64~2048 with almost equal VQ performance, compared with the universal codebook generation algorithm.

Efficient binary search space‐structured VQ encoder applied to a line spectral frequency quantisation in G.729 standard

This study presents a simple but high performance vector quantisation (VQ) codebook search algorithm for line spectral frequency quantisation in ITU-T G.729 speech codec, designated as the binary search space-structured VQ (BSS-VQ) algorithm. This is done through a combined use of a fast locating technique and lookup tables, such that an input vector, ahead of VQ encoding, is assigned efficiently to a subspace where merely a small number of codeword searches are required to be performed. As a consequence, the computational load is reduced remarkably. A trade-off can be made easy to meet a user's requirement when performing VQ encoding. It is experimentally validated that a search accuracy is well maintained at 99% approximately for a threshold of quantisation accuracy of 0.99. More importantly, with a full search algorithm as a benchmark for search load comparison, this proposal provides a 85% search load reduction, a figure far beyond 43% in multiple triangular inequality elimination approach, 47% in TIE with dynamic and intersection rules and 59% in quasi-binary search VQ algorithm.

Reversible data hiding in VQ index table with lossless coding and adaptive switching mechanism

In recent years, reversible data hiding (RDH) has attracted extensive attention from the academia. Besides RDH for uncompressed images, RDH for compressed images with satisfactory embedding performance is also important due to the universal use of image compression. In this paper, a RDH scheme for the encoded Vector Quantization (VQ) index table through the Improved Searching Order Coding (ISOC) is proposed. Original image is first compressed with a sorted VQ codebook and further losslessly encoded by the ISOC technique to produce an index table, in which all indices are categorized into three types, i.e., SOC type, Relative Addressing (RA) type and VQ type, based on the relationship between the index and its neighborhood. Each of the first two types of indices is adaptively switched or unchanged according to the current secret bit to be embedded, and the third type of indices isn't used to embed secret data. Experimental results show that, our scheme can embed secret data into 78.396% and 78.377% of all image blocks when VQ codebook sizes equal 256 and 1024, respectively. Compared with the reported schemes, our scheme can achieve higher hiding capacity with lower bit rate and extension degree of index table.

A New Compression Scheme Based on Adaptive Vector Quantization and Singular Value Decomposition

This paper presents a new technique for electrocardiogram data compression. The proposed method is based on the vector quantization and the singular value decomposition. First, the singular value decomposition is performed on the two-dimensional representation of the electrocardiogram signal and then a limited number of the weighted right singular vectors is quantized using the vector quantization. In addition, a residual encoding based on the singular value decomposition of the residual error is proposed. Consequently, a low reconstruction error was obtained. A new scheme is also proposed for the generation of the vector quantization codebook, it is based on the Lindth-Buzo-Gray algorithm, but in the proposed technique, at each vector quantization stage, a new codebook is generated. This latter is only produced from the retained singular vectors. Several tests were conducted using arrhythmia database of the Massachusetts Institute of Technology-Beth Israel Hospital. The obtained results were very satisfactory. Unlike most existing methods, in the proposed scheme, the codebook storage was taken into account during the calculation of the compression ratio which was not affected, since high compression ratios were obtained. For instance, a compression ratio of 102.54 was obtained in the case of the percent root mean square difference of 1.02%.

Vector Quantization for Satellite Image Compression

Image compression is the process of reducing the size of a file without humiliating the quality of the image to an unacceptable level by Human Visual System. The reduction in file size allows as to store more data in less memory and speed up the transmission process in low bandwidth also, in case of satellite images it reduces the time required for the image to reach the ground station. In order to increase the transmission process compression plays an important role in remote sensing images. This paper presents a coding scheme for satellite images using Vector Quantization. And it is a well-known technique for signal compression, and it is also the generalization of the scalar quantization. The given satellite image is compressed using VCDemo software by creating codebooks for vector quantization and the quality of the compressed and decompressed image is compared by the Mean Square Error, Signal to Noise Ratio, Peak Signal to Noise Ratio values.

PCA-Based Fast Search Method Using PCA-LBG-Based VQ Codebook for Codebook Search

A fast search method based on principle component analysis (PCA) is proposed to search codewords using vector quantization (VQ) codebooks obtained by PCA with Linde–Buzo–Gray (LBG) algorithms. The PCA sorts vectors of a test image and codewords of a PCA-LBG-based VQ codebook. The first search starts from the first codeword in the sorted codebook, and the next search starts from the previous best-matching codeword position in the sorted codebook. Both forward and backward searches are performed within the set search range until the best-matching codewords for all vectors of the test image are found in a sorted codebook. Because PCA efficiently distinguishes both test image vectors and codebook codewords, the proposed PCA-based fast search method outperforms the conventional algorithms in a codebook search. In particular, the experimental results show that, by using PCA-LBG-based VQ codebooks, the proposed PCA-based fast search method outperforms other methods in terms of peak signal-to-noise ratio for the compressed image, number of codewords searched, and runtime.

Modified firefly algorithm applied to image vector quantisation codebook design

Vector quantisation (VQ) has been used in signal processing applications, such as steganography, watermarking and signal compression. The performance of signal processing systems based on VQ depends on the designed codebooks. In the present work, modifications are introduced in the firefly algorithm - Linde-Buzo-Gray (FA-LBG) for the purpose of VQ codebook design. The modified version differs from the original one mainly for attempting to assure a stronger influence of the training set in codebook design. In the scenario of image vector quantisation, simulation results show that the modified algorithm outperforms both the original one as well as the traditional LBG algorithm in terms of codebook quality, assessed by the peak signal to noise ratio of the reconstructed images. Additionally, alternatives are presented to accelerate the proposed algorithm, which lead to execution time savings up to 90%.

Modified firefly algorithm applied to image vector quantisation codebook design

Vector quantisation VQ has been used in signal processing applications, such as steganography, watermarking and signal compression. The performance of signal processing systems based on VQ depends on the designed codebooks. In the present work, modifications are introduced in the firefly algorithm - Linde-Buzo-Gray FA-LBG for the purpose of VQ codebook design. The modified version differs from the original one mainly for attempting to assure a stronger influence of the training set in codebook design. In the scenario of image vector quantisation, simulation results show that the modified algorithm outperforms both the original one as well as the traditional LBG algorithm in terms of codebook quality, assessed by the peak signal to noise ratio of the reconstructed images. Additionally, alternatives are presented to accelerate the proposed algorithm, which lead to execution time savings up to 90%.

PSO Algorithm Applied to Codebook Design for Channel-Optimized Vector Quantization

Vector quantization (VQ) has been used in signal compression systems. However, in the scenario of image transmission, VQ is very sensitive to channel errors. An approach to decrease such sensitivity is channel-optimized vector quantization (COVQ), which involves VQ codebook design taking into account the characteristics of the channel. In the present work, particle swarm optimization (PSO) is applied to codebook design for COVQ. Simulation results are presented for a variety of bit error rates of a binary symmetric channel (BSC) and reveal the effectiveness of the method in decreasing visual impairment by blocking artifacts in the reconstructed images, overperforming conventional COVQ codebook design in terms of peak signal to noise ratio of the reconstructed images for approximately 90% of exhaustive evaluations of image transmission over BSC.

Very efficient variable-length codes for the lossless compression of VQ indices

In this paper, we propose a novel compression method that can efficiently compress a vector quantization (VQ) index table. Before compressing the VQ index table, the method sorts all of the codewords in the VQ codebook by principal component analysis (PCA), assuring that each codeword has extreme similarity to its adjacent codewords. Afterwards, in the VQ index table, the difference between the current compressed VQ index and one of its adjacent VQ indices is calculated as the compression code, and the indicators generated by the Huffman code method are used to identify the encoding length of each difference. In other words, each VQ index is replaced by one indicator and the difference, which are variable-length codes. The experimental results showed that the compression efficiency of the proposed method is superior to that of the other lossless data compression methods.

Performance comparisons between PCA-EA-LBG and PCA-LBG-EA approaches in VQ codebook generation for image compression

The aim of this study is to generate vector quantisation (VQ) codebooks by integrating principle component analysis (PCA) algorithm, Linde-Buzo-Gray (LBG) algorithm, and evolutionary algorithms (EAs). The EAs include genetic algorithm (GA), particle swarm optimisation (PSO), honey bee mating optimisation (HBMO), and firefly algorithm (FF). The study is to provide performance comparisons between PCA-EA-LBG and PCA-LBG-EA approaches. The PCA-EA-LBG approaches contain PCA-GA-LBG, PCA-PSO-LBG, PCA-HBMO-LBG, and PCA-FF-LBG, while the PCA-LBG-EA approaches contain PCA-LBG, PCA-LBG-GA, PCA-LBG-PSO, PCA-LBG-HBMO, and PCA-LBG-FF. All training vectors of test images are grouped according to PCA. The PCA-EA-LBG used the vectors grouped by PCA as initial individuals, and the best solution gained by the EAs was given for LBG to discover a codebook. The PCA-LBG approach is to use the PCA to select vectors as initial individuals for LBG to find a codebook. The PCA-LBG-EA used the final result of PCA-LBG as an initial individual for EAs to find a codebook. The search schemes in PCA-EA-LBG first used global search and then applied local search skill, while in PCA-LBG-EA first used local search and then employed global search skill. The results verify that the PCA-EA-LBG indeed gain superior results compared to the PCA-LBG-EA, because the PCA-EA-LBG explores a global area to find a solution, and then exploits a better one from the local area of the solution. Furthermore the proposed PCA-EA-LBG approaches in designing VQ codebooks outperform existing approaches shown in the literature.

An efficient tree‐structured vector quantization using dynamic triangular inequality elimination

As applied to a vector quantization (VQ) codebook search, a combined version of a dynamic triangular inequality elimination (DTIE) and a tree‐structured VQ (TSVQ) algorithm, designated as the DTIE–TSVQ approach, is presented in this letter as an efficient way to reach the aim of search performance improvement by successive updating of the search scope and reduced search load through the DTIE algorithm. In this manner, this proposal features the combined advantages of a TIE and a TSVQ algorithm such that 100% search accuracy is rendered together with a remarkable reduction in computational load. At the end of this work, this proposal is validated as a superior algorithm over TIE and conventional TSVQ algorithms by a high computational load saving up to 97.45% when dealing with line spectral frequency (LSF) quantization in a G.729 speech codec and image VQ encodings. © 2014 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.

A NEW VECTOR QUANTIZATION APPROACH FOR DISCRETE HMM SPEECH RECOGNITION SYSTEM

In order to address accuracy issues of discrete Hidden Markov Models (HMMs), in this paper, a new vector quantization (VQ) approach is presented. This new VQ approach performs an optimal distribution of VQ codebook components on HMM states. This technique that we named the distributed vector quantization (DVQ) of hidden Markov models, succeeds in unifying acoustic micro-structure and phonetic macro-structure, when the estimation of HMM parameters is performed. The DVQ technique is implemented through two variants. The first variant uses the K-means algorithm (K-means-DVQ) to optimize the VQ, while the second variant exploits the benefits of the classification behavior of neural networks (NN-DVQ) for the same purpose. The proposed variants are compared with the HMMbased baseline system by experiments of specific Arabic consonants recognition. The results show that the distributed vector quantization technique increase the performance of the discrete HMM system.