Vector Quantization Technique Research Articles

Low-complexity fractal-based image compression using two-stage search strategy

A fractal image compression scheme is presented which incorporates the vector quantizer (VQ) to reduce the computational complexity in the encoding phase. The high-detail blocks of an input image are approximated by a fast fractal coding, and the low-detail blocks are encoded by the VQ technique. In fractal image coding, most of the time is spent on finding a close match between a range block and a large pool of domain blocks. We propose a two-stage search algorithm to reduce the search space of the domain pool. First, a fast and simple binary pattern matching algorithm is proposed to extract a small pool of candidate domains, and then the gray-level-matching algorithm is used to find the best match from the candidates. In addition, a variable block-size segmentation method is used in our coding system to further improve the image quality. It is shown that the average encoding time is short, and the average speedup factor of 45 is achieved, as compared to the full search fractal coding. Simulation results also show that the proposed scheme can achieve better image quality and higher compression ratio than other fast fractal-based coding schemes.

Identification of Mineral Phases on Basalt Surfaces by Imaging SIMS

A method for the identification of mineral phases on basalt surfaces utilizing secondary ion mass spectrometry (SIMS) with imaging capability is described. The goal of this work is to establish the use of imaging SIMS for characterization of the surface of basalt. The basalt surfaces were examined by interrogating the intact basalt (heterogeneous mix of mineral phases) as well as mineral phases that have been separated from the basalt samples. Mineral separates from the basalt were used to establish reference spectra for the specific mineral phases. Electron microprobe and X-ray photoelectron spectroscopy were used as supplemental techniques for providing additional characterization of the basalt. Mineral phases that make up the composition of the basalt were identified from single-ion images which were statistically grouped. The statistical grouping is performed by utilizing a program that employs a generalized learning vector quantization technique. Identification of the mineral phases on the basalt surface is achieved by comparing the mass spectra from the statistically grouped regions of the basalt to the mass spectral results from the mineral separates. The results of this work illustrate the potential for using imaging SIMS to study adsorption chemistry at the top surface of heterogeneous mineral samples.

Model-based vector quantization with application to remotely sensed image data

Model-based vector quantization (MVQ) is introduced here as a variant of vector quantization (VQ). MVQ has the asymmetrical computational properties of conventional VQ, but does not require the use of pregenerated codebooks. This is a great advantage, since codebook generation is usually a computationally intensive process, and maintenance of codebooks for coding and decoding can pose difficulties. MVQ uses a simple mathematical model for mean removed errors combined with a human visual system model to generate parameterized codebooks. The error model parameter (lambda) is included with the compressed image as side information from which the same codebook is regenerated for decoding. As far as the user is concerned, MVQ is a codebookless VQ variant. After a brief introduction, the problems associated with codebook generation and maintenance are discussed. We then give a description of the MVQ algorithm, followed by an evaluation of the performance of MVQ on remotely sensed image data sets from NASA sources. The results obtained with MVQ are compared with other VQ techniques and JPEG/DCT. Finally, we demonstrate the performance of MVQ as a part of a progressive compression system suitable for use in an image archival and distribution installation.

A strategy for satellite data archival. Low noise variable-rate vector quantization with application to AVHRR satellite images: A tutorial review

Abstract The volume of satellite data amassed by modern day weather and climate satellites is so enormous that it has become virtually impossible for researchers to access the original resolution data collected by the satellites. Typically, researchers are forced to deal with lower resolution reduced data (e.g. cloud cover or temperatures) and often at a resolution degraded by one to three orders of magnitude when compared with the resolution of the original data. The uncertainties in the reduced data are often unknown. This state of affairs will only get worse, since the data to be collected under the guidance of NASA’s global change program may increase by several orders of magnitude in the coming decades. We need to experiment with mathematically rigorous ways to tame the original data without significantly degrading the information content. Compression of original data by objective mathematical techniques is a promising approach. This study adopts recent compression techniques developed in the field of communications and applies them to weather satellite data. Typically, these techniques compress the raw data by factors ranging from 10 to 100. The compressed data can be decompressed to retrieve the near original data at the site of the user. The mean error in the compression–decompression process varies from a few percent to several percent. As a demonstration, we consider the advanced very high resolution radiometer (AVHRR) radiances with a nadir resolution of 1 km ×1 km. For this demonstration, we adopt a well understood compression technique which is the so-called vector quantization technique. Several vector quantization techniques (full-search, tree-search, pruned-balanced-tree, greedy-tree and pruned-greedy-tree) are compared in performance and ease of implementation. The discussion focuses on the pruned greedy tree-structured vector quantizer because it is highly suited to the compression of AVHRR satellite images. For the case considered here, visual and scientific reproducibility of the original high resolution images are very good. The rms error for roughly 95% of the pixels in a scene is within 2%, even at a 32 : 1 compression ratio. The error in spatially averaged fields is less than 0.1% for averaging scales in excess of 50 km ×50 km. Some important spatial structural information is lost, however. It is found that the same image when compressed using JPEG standard shows significant loss of numerical accuracy at the same compression ratio of 32 : 1. But improvements and developments in compression techniques can minimize these errors and afford researchers the luxury of storing and working with high resolution data at roughly at about 0.03 of the space required by the original data.

Damage characterization of carbon/carbon laminates using neural network techniques on AE signals

Carbon/Carbon (C/C) composites have a large number of microcracks in matrix-matrix or fiber-matrix regions resulting from the thermal processes during manufacturing. Although not harmful to the overall structural integrity, such a network of microcracks creates a noisy background from early load application which ‘covers’ the onset of critical failure mechanisms. Conventional AE analysis based on a sudden activity increase observed in the cumulative events versus load plot, or amplitude distribution histograms, provides limited information on this type of AE activity. Instead, multivariate techniques of unsupervised pattern recognition, taking into account a large number of AE signal descriptors, are proven useful for the clustering of similar AE events. AE results from a systematic fracture mechanics study of 2D woven C/C laminates are analyzed in this paper. Artificial Neural System (ANS) methods are employed for the clustering of similar AE signals, enabling a phenomenological correlation with the actual failure modes. The numerical procedure introduces a modified Learning Vector Quantization (LVQ) technique which was proved fast and suitable for the type of AE data emitted by composites. Fractographic evidence from failed tensile coupons corroborates the predictions of the numerical method in recognizing different failure mechanisms. Cumulative event charts of the various classes versus load demonstrate the criticality of each class on final coupon failure, and may lead to the definition of reliable criteria for the evaluation of remaining strength or life.

Two-dimensional split and merge algorithm for differential vector quantization of images

Abstract In this paper, an image is modelled as a two and a half-dimensional surface and an approximated surface for the image is formed by triangular patches. A new two-dimensional split and merge algorithm (2DSM) for generating the approximated surface has been devised. The algorithm iteratively improves the approximated surface by splitting and merging the triangles in order to drive the error under a specified bound. In addition, a new optimal triangulation method for image data approximation is proposed. The algorithm is successfully applied for coding of monochrome images using differential vector quantization (DVQ) technique. Simulation results show that the proposed method can achieve 2.8 dB improvement on the approximated image and 0.68 dB improvement on the decoded image at a bit-rate lower than the current schemes. Besides, excellent reconstruction visual quality is observed.

Frame-Adaptive Vector Quantization

Vector quantization (VQ) is a promising technique for low bit rate image compression. Recently image sequence compression algorithms based on VQ have been reported in the literature. Image sequences are highly nonstationary and generally exhibit variations from frame to frame and from scene to scene; hence using a fixed VQ codebook to encode the different frames/sequences may not always guarantee a good coding performance. Several adaptive techniques which improve the coding performance have been reported in the literature. However, we note that most adaptive techniques result in further increases in computational complexity and/or the bit rate. In this paper, a new frame adaptive VQ technique for image sequence compression (FAVQ) is presented. This technique exploits the inter/intraframe correlations and provides frame adaptability at a reduced complexity. A dynamic self organized codebook is used to track the local statistics from frame to frame. Computer simulations using standard CCITT sequences demonstrate the superior coding performance of FAVQ. In addition, FAVQ is a single pass technique which makes possible real time implementation.

Time-invariant and time-varying multirate filter banks : application to image coding

This paper reviews various concepts and solutions of time-invariant and time-varying multirate filter banks. It discusses their performance for image and video coding at low bit rates, and their applicability in the mpeg-4 framework. Time-invariant multirate filter banks, and methods of design with different criteria appropriate for signal compression are first presented. Several procedures of quantization, namely scalar and lattice vector quantization, with bit allocation optimized in the rate-distortion sense, are used for the encoding of the subband signals. A technique of rate-constrained lattice vector quantization (rc-lvq), combined with a three components entropy coding, allow, together with distortion psychovi-sual weighting mechanisms to obtain significant visual improvements versus scalar quantization or the zerotree technique. However, time-invariant multirate filter banks, although efficient in terms of compression, are not well suited for content-based functionalities. Content-based features may require the ability to manipulate and thus encode a given region in the scene independently of the neighbouring regions, hence the use of transformations that can be adapted to arbitrary size bounded supports. Also, to increase the compression efficiency, one may want to adapt the transformation to the region characteristics, and thus use transform switching mechanisms, with soft or hard transitions. Three main classes of transformations can address these problems: shape-adaptive block transforms, transforms relying on signal extensions and transforms relying on time-varying multirate filter banks. These various solutions, with their methods of design, are reviewed. Emphasis is put on an extension of the SDF (symmetric delay factorization) technique which opens new perspectives in the design of time-bounded and time-varying filter banks. A region-adapted rate-distortion quantization algorithm has been used in the evaluation of the transformations compression efficiency. The coding results illustrate the interest of these techniques for compression but also for features such as quality scalability applied to selected regions of the image.

Classified vector quantisation using principal components

The authors describe a novel classified vector quantisation technique that uses a principal component-based classifier. Each non-shade class is associated with a vector representing the largest principal component of the training vectors for that class. The technique works directly in the spatial domain without any preprocessing and achieves good perceptual quality and edge integrity at bit rates well below 1 bit/pixel with reduced coding complexity.

Hybrid block truncation coding

A hybrid block truncation coding (BTC) is presented. In the hybrid BTC, a universal codebook using Hamming codes and a differential pulse code modulation (DPCM) are employed, respectively, to the bit plane and the side information of BTC to reduce coding rate. Simulation results reveal that the performance of the proposed algorithm is only slightly worse than that of the hybrid BTC using vector quantization (VQ) techniques, but with much lower computational or hardware complexity.

A novel interpolative vector quantization scheme

Abstract A novel adaptive interpolative vector quantization scheme is developed and presented in this paper. Instead of constructing an approximation of the original image with the traditional fix-rate subsampling and interpolation process, a new dominant point detection algorithm is employed to build a non-uniformly spaced decimation lattice. In effect, better approximation and lower bit-rate are achieved by extracting only those sampling points which have significant effects on the aliasing error. The difference between the original and the approximated image is encoded with vector quantization . Experimental results show that the decoded images are found to exhibit coding fidelity of over 34 dB with improved visual quality over existing vector quantization techniques. A comparison of the proposed scheme with other existing methods of similar complexity has also been presented.

Radar signal interpretation using neural network for defect detection in concrete : Shourky, S.N.; Martinelli, D.; Varadarajan, S.T.; Halabe, U.B. Materials Evaluation, Vol. 54, No. 3, pp. 393–397 (Mar. 1996)

A neural network based on the learning vector quantization (LVQ) technique is applied for the classification of radar echo waveforms obtained from concrete bridge deck specimens with varying internal defects. A randomly chosen set of as measured (raw) time-domain echosignals was used in training the network to classify waveforms into those representing defective and nondefective conditions in the concrete specimens. The same set of waveforms as preprocessed by subtracting the reference noise signal and was again separately used to train the networks. Different versions of LVQ algorithms were used independently in training, and the best results in terms of the recognition accuracy were obtained by training using the OLVQ (optimized LVQ) algorithm. The OLVQ trained neural network was able to classify an independent set of test waveforms with a higher degree of accuracy than those trained with other variants of LVQ algorithms. It has been observed that the subtraction of the reference noise component (preprocessing of the time domain signal) had no effect on the performance of the correct recognition of the OLVQ network. The performance of the network indicates its high potential for use in the analysis and interpretation of raw radar echo waveforms. This technique also holds promise for successful classification of radar waveforms into classes that represent specific kind of internal defects in concrete specimens such as cracks, delaminations, voids, etc.

Voronoi Projection-Based Fast Nearest-Neighbor Search Algorithms: Box-Search and Mapping Table-Based Search Techniques

In this paper we consider fast nearest-neighbor search techniques based on the projections of Voronoi regions. The Voronoi diagram of a given set of points provides an implicit geometric interpretation of nearest-neighbor search and serves as an important basis for several proximity search algorithms in computational geometry and in developing structure-based fast vector quantization techniques. The Voronoi projections provide an approximate characterization of the Voronoi regions with respect to their locus property of localizing the search to a small subset of codevectors. This can be viewed as a simplified and practically viable equivalent of point location using the Voronoi diagram while circumventing the complexity of the full Voronoi diagram. In this paper, we provide a comprehensive study of two fast search techniques using the Voronoi projections, namely, the box-search and mapping table-based search in the context of vector quantization encoding. We also propose and study the effect and advantage of using the principal component axes for data with high degree of correlation across their components, in reducing the complexity of the search based on Voronoi projections.

Nonparametric classifier design using greedy tree-structured vector quantization technique

In this paper, we propose a novel tree-structured vector quantization (TSVQ) design algorithm for the applications of nonparametric pattern recognition. The TSVQ design algorithm is used to reduce the large size of the design sets required by a nonparametric classifier. For an N-class problem, the TSVQ consists of N branches with one for each class. Using the design sets as training data, the algorithm splits the leaf nodes in a greedy manner to minimize the classification error rate for tree-growing. Simulation results show that the classifiers designed using this new algorithm require less classification time than that required by other design set reduction algorithms. In addition, in many cases, the new classifiers enjoy almost the same low error rate as that of traditional k-NN nonparametric classifiers.

Variable-rate CELP based on subband flatness

Code-excited linear prediction (CELP) is the predominant methodology for communications quality speech coding below 8 kbps, and several variable-rate CELP schemes have been discussed in the literature, including QCELP, the variable-rate wideband digital cellular mobile radio speech coding standard specified in IS-95. A key component of these speech coders is the detection and classification of speech activity, and several cues for rate variation have been studied, such as measuring the short-term speech energy, deciding whether the speech is voiced or unvoiced, or making more sophisticated phonetic classifications. We present a new method for rate variation based on a measure of subband spectral flatness, called spectral entropy. Spectral entropy is a normalized indicator of the texture of the input spectrum and is thus less dependent on speech and background noise energy variations. We present some results on the use of spectral entropy for voice activity detection across subbands and then evaluate using spectral entropy for deriving mode and rate allocation cues for a variable-rate CELP coder operating at an average rate of 2 kbps. To achieve communications quality speech at this rate, we develop a new split-band vector quantization (VQ) technique for representing the line spectral pairs and a multiple codebook approach for efficiently quantizing the coefficients of a three-tap pitch predictor, called lag-indexed VQ.

Fractal block coding: A new method

Fractal image compression[1] is a completely new method to compress images by searching and exploiting the self-similarity of the whole image[2]. Fractal Block Coding (FBC) is a practicable fractal coding scheme with annoying slow encoding speed[3]. In this paper, we classify the image blocks by Classified Vector Quantization (CVQ) technique and present an Adaptive Block Truncation Coding (ABTC) scheme to process the midrange blocks in the image. By this method, we reduce the encoding time to one forty-fifth comparing to ordinary FBC method with little change in compression ratio and a little decreased coded image quality.

Storage- and entropy-constrained multi-stage vector quantization and its applications to progressive image transmission

A new vector quantization technique, called storage- and entropy-constrained multi-stage vector quantization (SECMSVQ) technique, is presented for progressive image transmission (PIT) applications. In the SECMSVQ technique, the rate constraint at each stage of the vector quantizer (VQ), and the storage size of the whole VQ are specified independently before the design. A dynamic programming technique is then used to allocate the storage size to each stage of the VQ optimally so that the quantization distortion of the VQ is minimized. The SECMSVQ is well-suited to constructing a PIT system where the transmission time and storage size are the major concerns. Moreover, simulation results show that the SECMSVQ technique can be used effectively to implement signal compression systems under the storage and rate constraints.

Reducing blocking artifacts within vector quantization algorithms

Vector quantization (VQ) provides a fast, low-power image compression technique for video communication products. However, since VQ techniques operate on each block of the image separately, they have been criticized for producing visually-disturbing blocking artifacts, especially at low bit rates. We present a VQ technique which uses hexagonally-shaped vectors to greatly reduce blocking artifacts. The effectiveness of the technique is demonstrated experimentally.

Novel image compression method using edge-oriented classifier and novel predictive noiseless coding method

A new image compression approach is proposed in which variable block size technique is adopted, using quadtree decomposition, for coding images at low bit rates. In the proposed approach, low-activity regions, which usually occupy large areas in an image, were coded with a larger block size and the block mean is used to represent each pixel in the block. To preserve edge integrity, the classified vector quantisation (CVQ) technique is used to code high-activity regions. A new edge-oriented classifier without employing any thresholds is proposed for edge classification. A novel predictive noiseless coding (NPNC) method which exploits the redundancy between neighbouring blocks is also presented to efficiently code the mean values of low-activity blocks and the addresses of edge blocks. The bit rates required for coding the mean values and addresses can be significantly reduced by the proposed NPNC method. Experimental results show that excellent reconstructed images and higher PSNR were obtained.

Classified vector quantizer technique for image coding employing the linear phase paraunitary <italic>M</italic>‐band filter bank

We propose an efficient classified vector quantizer (CVQ) technique employing the linear phase paraunitary (LPPU) M-band filter bank. Based on the asymptotic distortion of VQ, a coding gain, which includes the gain by the correlation in the vector as well as the energy compaction, is first derived. Then we design the LPPU four-band and eight-band filter banks, with various lengths, that maximize the coding gain. Among the designed LPPU filter banks, we select the LPPU four- band filter bank for the proposed CVQ technique, since we can signifi- cantly alleviate the complexity of CVQ and classify the input image more accurately with 434 blocks than with 838 blocks. However, the side information for the 434 blocks is four times that for the 838 blocks. In order to reduce the side information, we propose the quadtree method, which reduces the side information by about 50%. Simulation results demonstrate that the proposed technique yields about 0.8- to 1.0-dB peak-SNR gain over that of the Joint Photographic Experts Group in the range of 0.3 to 0.4 bits per pixel. © 1996 Society of Photo-Optical Instrumentation Engineers.