Summary form only given. A simple and fast audio signal compression method that uses subband filtering and quantization is presented. The method is suitable for compression of telephone quality audio signals. It can compress four CCITT 64 kbit/s PCM A- or /spl mu/-coded speech channels into one channel with sufficient sound quality for telephone use. A straightforward implementation of the compression and decompression methods have the following steps. First the incoming speech signal is converted from a /spl mu/ or A-law coded signal into 16 bit linear PCM signal and then divided into 16 bands of equal bandwidth by using the analysis filter bank. Then the sampling frequencies of the frequency channels are decreased by a factor of 16. After this decimation the subband samples are fed to a fixed quantizer. Finally the quantized subband values and the side information needed for decoding is packed into a data stream and sent to the receiver.
只提供摘要形式。提出了一种基于子带滤波和量化的音频信号压缩方法。该方法适用于电话级音频信号的压缩。它可以将4个CCITT 64kbit /s PCM A或/spl mu/编码语音通道压缩成一个具有足够音质的电话通道。压缩和解压缩方法的简单实现有以下步骤。首先将输入的语音信号从a /spl mu/ or a -law编码信号转换为16位线性PCM信号，然后使用分析滤波器组将其划分为16个等带宽的频带。然后将频率通道的采样频率降低16倍。在抽取后，子带样本被送入固定的量化器。最后将量化后的子带值和解码所需的侧信息打包成数据流发送给接收端。
{"title":"Fast subband coder for telephone quality audio","authors":"H. Raittinen, K. Kaski","doi":"10.1109/DCC.1995.515581","DOIUrl":"https://doi.org/10.1109/DCC.1995.515581","url":null,"abstract":"Summary form only given. A simple and fast audio signal compression method that uses subband filtering and quantization is presented. The method is suitable for compression of telephone quality audio signals. It can compress four CCITT 64 kbit/s PCM A- or /spl mu/-coded speech channels into one channel with sufficient sound quality for telephone use. A straightforward implementation of the compression and decompression methods have the following steps. First the incoming speech signal is converted from a /spl mu/ or A-law coded signal into 16 bit linear PCM signal and then divided into 16 bands of equal bandwidth by using the analysis filter bank. Then the sampling frequencies of the frequency channels are decreased by a factor of 16. After this decimation the subband samples are fed to a fixed quantizer. Finally the quantized subband values and the side information needed for decoding is packed into a data stream and sent to the receiver.","PeriodicalId":107017,"journal":{"name":"Proceedings DCC '95 Data Compression Conference","volume":"62 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127098884","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Summary form only given. As part of an industry standardization effort, we have evaluated compression algorithms for throughput enhancement in a synchronous communication environment. Synchronous data compression systems are link layer compressors used between digital transmission devices in internetworks to increase effective throughput. Compression is capable of speeding such links, but achievable performance is effected by interaction of algorithm, the networking protocols, and implementation details. The compression environment is different from traditional file compression in inducing a trade-off between compression ratio, compression time, and the performance metric (network throughput). In addition, other parameters and behavior are introduced, including robustness to data retransmission and multiple interleaved streams. Specifically, we have evaluated the following issues through both synchronous queuing and direct network simulation: (1) relative algorithm capability; (2) throughput improvement for various algorithms as a function of compression processor capability; (3) the impact of multiple compression context; (4) protocol interactions; and (5) specialized algorithms.
{"title":"Algorithm evaluation for synchronous data compression","authors":"M.W. Maier","doi":"10.1109/DCC.1995.515554","DOIUrl":"https://doi.org/10.1109/DCC.1995.515554","url":null,"abstract":"Summary form only given. As part of an industry standardization effort, we have evaluated compression algorithms for throughput enhancement in a synchronous communication environment. Synchronous data compression systems are link layer compressors used between digital transmission devices in internetworks to increase effective throughput. Compression is capable of speeding such links, but achievable performance is effected by interaction of algorithm, the networking protocols, and implementation details. The compression environment is different from traditional file compression in inducing a trade-off between compression ratio, compression time, and the performance metric (network throughput). In addition, other parameters and behavior are introduced, including robustness to data retransmission and multiple interleaved streams. Specifically, we have evaluated the following issues through both synchronous queuing and direct network simulation: (1) relative algorithm capability; (2) throughput improvement for various algorithms as a function of compression processor capability; (3) the impact of multiple compression context; (4) protocol interactions; and (5) specialized algorithms.","PeriodicalId":107017,"journal":{"name":"Proceedings DCC '95 Data Compression Conference","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127883378","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Summary form only given, substantially as follows. Transform coded images suffer from specific image degradations. In the case of standard JPEG compression/decompression the image quality losses are known to be blocking effects resulting from mean value discontinuities along the 8*8 pixel block boundaries as well as ringing artifacts due to the limited precision of the reconstruction from linear combinations of quantized or discarded basis functions. The most evident consequence of JPEG compression is the fragmentation of image histograms mainly caused by blocking in low activity image subareas. The histogram of the image shows spikes that contain most of the signal amplitudes, the other values are distributed on the remaining permissible levels. As a measure of the blocking effect, the blocking factor is defined as the ratio of the spikes area to the total area of image histogram. This method represents a promising approach to the control of locally adaptive image deblocking when the necessary enhancement depends on the local image characteristics. The blocking factor is easy to compute and provides a direct measure of the local image degradation due to blocking. A new deblocking algorithm is proposed.
{"title":"Histogram analysis of JPEG compressed images as an aid in image deblocking","authors":"M. Datcu, G. Schwarz, K. Schmidt, C. Reck","doi":"10.1109/DCC.1995.515535","DOIUrl":"https://doi.org/10.1109/DCC.1995.515535","url":null,"abstract":"Summary form only given, substantially as follows. Transform coded images suffer from specific image degradations. In the case of standard JPEG compression/decompression the image quality losses are known to be blocking effects resulting from mean value discontinuities along the 8*8 pixel block boundaries as well as ringing artifacts due to the limited precision of the reconstruction from linear combinations of quantized or discarded basis functions. The most evident consequence of JPEG compression is the fragmentation of image histograms mainly caused by blocking in low activity image subareas. The histogram of the image shows spikes that contain most of the signal amplitudes, the other values are distributed on the remaining permissible levels. As a measure of the blocking effect, the blocking factor is defined as the ratio of the spikes area to the total area of image histogram. This method represents a promising approach to the control of locally adaptive image deblocking when the necessary enhancement depends on the local image characteristics. The blocking factor is easy to compute and provides a direct measure of the local image degradation due to blocking. A new deblocking algorithm is proposed.","PeriodicalId":107017,"journal":{"name":"Proceedings DCC '95 Data Compression Conference","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131498430","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
We propose a predictive coding algorithm for lossy compression of digital halftones produced by clustered-dot dithering. In our scheme, the predictor estimates the size and shape of each halftone dot (cluster) based on the characteristics of neighboring clusters. The prediction template depends on which portion, or sub-cell, of the dithering matrix produced the dot. Information loss is permitted through imperfect representation of the prediction residuals. For some clusters, no residual is transmitted at all, and for others, information about the spatial locations of bit errors is omitted. Specifying only the number of bit errors in the residual is enough to allow the decoder to form an excellent approximation to the original dot structure. We also propose a simple alternative to the ordinary Hamming distance for computing distortion in bi-level images. Experiments with 1024/spl times/1024 images, 8/spl times/8 dithering cells, and 600 dpi printing have shown that the coding algorithm maintains good image quality while achieving rates below 0.1 bits per pixel.
{"title":"Lossy compression of clustered-dot halftones using sub-cell prediction","authors":"R. A. V. Kam, R. Gray","doi":"10.1109/DCC.1995.515501","DOIUrl":"https://doi.org/10.1109/DCC.1995.515501","url":null,"abstract":"We propose a predictive coding algorithm for lossy compression of digital halftones produced by clustered-dot dithering. In our scheme, the predictor estimates the size and shape of each halftone dot (cluster) based on the characteristics of neighboring clusters. The prediction template depends on which portion, or sub-cell, of the dithering matrix produced the dot. Information loss is permitted through imperfect representation of the prediction residuals. For some clusters, no residual is transmitted at all, and for others, information about the spatial locations of bit errors is omitted. Specifying only the number of bit errors in the residual is enough to allow the decoder to form an excellent approximation to the original dot structure. We also propose a simple alternative to the ordinary Hamming distance for computing distortion in bi-level images. Experiments with 1024/spl times/1024 images, 8/spl times/8 dithering cells, and 600 dpi printing have shown that the coding algorithm maintains good image quality while achieving rates below 0.1 bits per pixel.","PeriodicalId":107017,"journal":{"name":"Proceedings DCC '95 Data Compression Conference","volume":"70 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126915873","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Summary form only given, as follows. Ray-casting, that is, calculating the intersections of a large array of lines with a solid object is a well-known technique that is central to many algorithms useful in solid modeling. Ray-casting is a compact and elegant way for displaying and calculating the geometrical properties of 3-D objects. The Ray-Casting Engine RCE-1.5 is an application specific massively parallel computer dedicated to ray-casting 3D objects. We present an application specific hardware-oriented data compression algorithm. We developed a simple yet powerful data compression hardware specifically tailored to compressing ray-files, the data structure internal to the RCE-1.5. We have used the compression hardware to meet performance goals while reducing the cost of building the RCE-1.5. We had to balance compression performance on the one hand with real time constraints, development time constraints and hardware costs on the other. With a modest amount of compression hardware we were able to more than double the internal and external data transfer rates. In addition we more than doubled the effective internal memory buffer size. The increase throughput rate enabled us to use (slow but inexpensive) DRAM rather than (faster but expensive) SRAM, dramatically reducing the over-all system cost. This is but one example where judicious use of data compression techniques can dramatically enhance system performance while at the same time reducing the system cost.
{"title":"Application specific hardware compression of ray-casting data","authors":"G. Kedem, T. Alexander","doi":"10.1109/DCC.1995.515594","DOIUrl":"https://doi.org/10.1109/DCC.1995.515594","url":null,"abstract":"Summary form only given, as follows. Ray-casting, that is, calculating the intersections of a large array of lines with a solid object is a well-known technique that is central to many algorithms useful in solid modeling. Ray-casting is a compact and elegant way for displaying and calculating the geometrical properties of 3-D objects. The Ray-Casting Engine RCE-1.5 is an application specific massively parallel computer dedicated to ray-casting 3D objects. We present an application specific hardware-oriented data compression algorithm. We developed a simple yet powerful data compression hardware specifically tailored to compressing ray-files, the data structure internal to the RCE-1.5. We have used the compression hardware to meet performance goals while reducing the cost of building the RCE-1.5. We had to balance compression performance on the one hand with real time constraints, development time constraints and hardware costs on the other. With a modest amount of compression hardware we were able to more than double the internal and external data transfer rates. In addition we more than doubled the effective internal memory buffer size. The increase throughput rate enabled us to use (slow but inexpensive) DRAM rather than (faster but expensive) SRAM, dramatically reducing the over-all system cost. This is but one example where judicious use of data compression techniques can dramatically enhance system performance while at the same time reducing the system cost.","PeriodicalId":107017,"journal":{"name":"Proceedings DCC '95 Data Compression Conference","volume":"48 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125371977","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Summary form only given. A direct sum codebook (DSC) has the potential to reduce both memory and computational costs of vector quantization. A DSC consists of several sets or stages of vectors. An equivalent code vector is made from the direct sum of one vector from each stage. Such a structure, with p stages containing m vectors each, has m/sup p/ equivalent code vectors, while requiring the storage of only mp vectors. DSC quantizers are not only memory efficient, they also have a naturally simple encoding algorithm, called a residual encoding. A residual encoding uses the nearest neighbor at each stage, requiring comparison with mp vectors rather than all m/sup p/ possible combinations. Unfortunately, this encoding algorithm is suboptimal because of a problem called entanglement. Entanglement occurs when a different vector from that obtained by a residual encoding is actually a better fit for the input vector. An optimal encoding can be obtained by an exhaustive search, but this sacrifices the savings in computation. Lattice-based DSC quantizers are designed to be optimal under a residual encoding by avoiding entanglement Successive stages of the codebook produce finer and finer partitions of the space, resulting in equivalent code vectors which are points in a truncated lattice. After the initial design, the codebook can be optimized for a given source, increasing performance beyond that of a simple lattice vector quantizer. Experimental results show that DSC quantizers based on cubical lattices perform as well as exhaustive search quantizers on a scalar source.
{"title":"Lattice-based designs of direct sum codebooks for vector quantization","authors":"C. Barrett, R. L. Frost","doi":"10.1109/DCC.1995.515546","DOIUrl":"https://doi.org/10.1109/DCC.1995.515546","url":null,"abstract":"Summary form only given. A direct sum codebook (DSC) has the potential to reduce both memory and computational costs of vector quantization. A DSC consists of several sets or stages of vectors. An equivalent code vector is made from the direct sum of one vector from each stage. Such a structure, with p stages containing m vectors each, has m/sup p/ equivalent code vectors, while requiring the storage of only mp vectors. DSC quantizers are not only memory efficient, they also have a naturally simple encoding algorithm, called a residual encoding. A residual encoding uses the nearest neighbor at each stage, requiring comparison with mp vectors rather than all m/sup p/ possible combinations. Unfortunately, this encoding algorithm is suboptimal because of a problem called entanglement. Entanglement occurs when a different vector from that obtained by a residual encoding is actually a better fit for the input vector. An optimal encoding can be obtained by an exhaustive search, but this sacrifices the savings in computation. Lattice-based DSC quantizers are designed to be optimal under a residual encoding by avoiding entanglement Successive stages of the codebook produce finer and finer partitions of the space, resulting in equivalent code vectors which are points in a truncated lattice. After the initial design, the codebook can be optimized for a given source, increasing performance beyond that of a simple lattice vector quantizer. Experimental results show that DSC quantizers based on cubical lattices perform as well as exhaustive search quantizers on a scalar source.","PeriodicalId":107017,"journal":{"name":"Proceedings DCC '95 Data Compression Conference","volume":"49 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114582108","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
We consider sublinear massively parallel algorithms for compressing text with respect to a static dictionary. Algorithms for the PRAM model can do this optimally in O(m+log(n)) time with n processors, where m is the length of the longest entry in the dictionary and n is the length of the input string. We consider what is perhaps the most practical model of massively parallel computation imaginable: a linear array of processors where each processor is connected only to its left and right neighbors. We present an algorithm which in time O(km+mlog(m)) with n/(km) processors is guaranteed to be within a factor of (k+1)/k of optimal, for any integer k/spl ges/1. We also present experiments indicating that performance may be even better in practice.
{"title":"Near optimal compression with respect to a static dictionary on a practical massively parallel architecture","authors":"D. Belinskaya, S. Agostino, J. Storer","doi":"10.1109/DCC.1995.515507","DOIUrl":"https://doi.org/10.1109/DCC.1995.515507","url":null,"abstract":"We consider sublinear massively parallel algorithms for compressing text with respect to a static dictionary. Algorithms for the PRAM model can do this optimally in O(m+log(n)) time with n processors, where m is the length of the longest entry in the dictionary and n is the length of the input string. We consider what is perhaps the most practical model of massively parallel computation imaginable: a linear array of processors where each processor is connected only to its left and right neighbors. We present an algorithm which in time O(km+mlog(m)) with n/(km) processors is guaranteed to be within a factor of (k+1)/k of optimal, for any integer k/spl ges/1. We also present experiments indicating that performance may be even better in practice.","PeriodicalId":107017,"journal":{"name":"Proceedings DCC '95 Data Compression Conference","volume":"50 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123317271","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Summary form only given. General-purpose text compression works normally at the lexical level assuming that symbols to be encoded are independent or they depend on preceding symbols within a fixed distance. Traditionally such syntactical models have been focused on compression of source programs, but also other areas are feasible. The compression of a parse tree is an important and challenging part of syntactical modeling. A parse tree can be represented by a left parse which is a sequence of productions applied in preorder. A left parse can be encoded efficiently with arithmetic coding using counts of production alternatives of each nonterminal. We introduce a more refined method which reduces the size of a compressed tree. A blending scheme, PPM (prediction by partial matching) produces very good compression on text files. In PPM, adaptive models of several context lengths are maintained and they are blended during processing. The k preceding symbols of the symbol to be encoded form the context of order k. We apply the PPM technique to a left parse so that we use contexts of nodes instead of contexts consisting of preceding symbols in the sequence. We tested our approach with parse trees of Pascal programs. Our method gave on the average 20 percent better compression than the standard method based on counts of production alternatives of nonterminals. In our model, an item of the context is a pair (production, branch). The form of the item seems to be crucial. We tested three other variations for an item: production, nonterminal, and (nonterminal, branch), but all these three approaches produced clearly worse results.
{"title":"Context coding of parse trees","authors":"J. Tarhio","doi":"10.1109/DCC.1995.515552","DOIUrl":"https://doi.org/10.1109/DCC.1995.515552","url":null,"abstract":"Summary form only given. General-purpose text compression works normally at the lexical level assuming that symbols to be encoded are independent or they depend on preceding symbols within a fixed distance. Traditionally such syntactical models have been focused on compression of source programs, but also other areas are feasible. The compression of a parse tree is an important and challenging part of syntactical modeling. A parse tree can be represented by a left parse which is a sequence of productions applied in preorder. A left parse can be encoded efficiently with arithmetic coding using counts of production alternatives of each nonterminal. We introduce a more refined method which reduces the size of a compressed tree. A blending scheme, PPM (prediction by partial matching) produces very good compression on text files. In PPM, adaptive models of several context lengths are maintained and they are blended during processing. The k preceding symbols of the symbol to be encoded form the context of order k. We apply the PPM technique to a left parse so that we use contexts of nodes instead of contexts consisting of preceding symbols in the sequence. We tested our approach with parse trees of Pascal programs. Our method gave on the average 20 percent better compression than the standard method based on counts of production alternatives of nonterminals. In our model, an item of the context is a pair (production, branch). The form of the item seems to be crucial. We tested three other variations for an item: production, nonterminal, and (nonterminal, branch), but all these three approaches produced clearly worse results.","PeriodicalId":107017,"journal":{"name":"Proceedings DCC '95 Data Compression Conference","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123591496","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The authors have proposed a compressed data format that can be used with any foreign file system and that allows users to access data randomly in a compressed file without entirely decompressing it. Since the compressed file includes all information regarding compression in this format, there is a great advantage that any file system can treat compressed files as just usual files, even if the file system does not have compression capability.
{"title":"A universal compressed data format for foreign file systems","authors":"T. Kawashima, T. Igarashi, R. Hines, M. Ogawa","doi":"10.1109/DCC.1995.515539","DOIUrl":"https://doi.org/10.1109/DCC.1995.515539","url":null,"abstract":"The authors have proposed a compressed data format that can be used with any foreign file system and that allows users to access data randomly in a compressed file without entirely decompressing it. Since the compressed file includes all information regarding compression in this format, there is a great advantage that any file system can treat compressed files as just usual files, even if the file system does not have compression capability.","PeriodicalId":107017,"journal":{"name":"Proceedings DCC '95 Data Compression Conference","volume":"94 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129412363","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Previous works, including adaptive quantizer selection and adaptive coefficient thresholding, have addressed the optimization of a baseline-decodable JPEG coder in a rate-distortion (R-D) sense. In this work, by developing an entropy-constrained quantization framework, we show that these previous works do not fully realize the attainable coding gain, and then formulate a computationally efficient way that attempts to fully realize this gain for baseline-JPEG-decodable systems. Interestingly, we find that the gains obtained using the previous algorithms are almost additive. The framework involves viewing a scalar-quantized system with fixed quantizers as a special type of vector quantizer (VQ), and then to use techniques akin to entropy-constrained vector quantization (ECVQ) to optimize the system. In the JPEG case, a computationally efficient algorithm can be derived, without training, by jointly performing coefficient thresholding, quantizer selection, and Huffman table customization, all compatible with the baseline JPEG syntax. Our algorithm achieves significant R-D improvement over standard JPEG (about 2 dB for typical images) with performance comparable to that of more complex "state-of-the-art" coders. For example, for the Lenna image coded at 1.0 bits per pixel, our JPEG-compatible coder achieves a PSNR of 39.6 dB, which even slightly exceeds the published performance of Shapiro's wavelet coder. Although PSNR does not guarantee subjective performance, our algorithm can be applied with a flexible range of visually-based distortion metrics.
{"title":"JPEG optimization using an entropy-constrained quantization framework","authors":"M. Crouse, K. Ramchandran","doi":"10.1109/DCC.1995.515524","DOIUrl":"https://doi.org/10.1109/DCC.1995.515524","url":null,"abstract":"Previous works, including adaptive quantizer selection and adaptive coefficient thresholding, have addressed the optimization of a baseline-decodable JPEG coder in a rate-distortion (R-D) sense. In this work, by developing an entropy-constrained quantization framework, we show that these previous works do not fully realize the attainable coding gain, and then formulate a computationally efficient way that attempts to fully realize this gain for baseline-JPEG-decodable systems. Interestingly, we find that the gains obtained using the previous algorithms are almost additive. The framework involves viewing a scalar-quantized system with fixed quantizers as a special type of vector quantizer (VQ), and then to use techniques akin to entropy-constrained vector quantization (ECVQ) to optimize the system. In the JPEG case, a computationally efficient algorithm can be derived, without training, by jointly performing coefficient thresholding, quantizer selection, and Huffman table customization, all compatible with the baseline JPEG syntax. Our algorithm achieves significant R-D improvement over standard JPEG (about 2 dB for typical images) with performance comparable to that of more complex \"state-of-the-art\" coders. For example, for the Lenna image coded at 1.0 bits per pixel, our JPEG-compatible coder achieves a PSNR of 39.6 dB, which even slightly exceeds the published performance of Shapiro's wavelet coder. Although PSNR does not guarantee subjective performance, our algorithm can be applied with a flexible range of visually-based distortion metrics.","PeriodicalId":107017,"journal":{"name":"Proceedings DCC '95 Data Compression Conference","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129883739","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：