Summary form only given. Error resilience is an important requirement when errors occur during video transmission. The video transmitted over the Internet is usually a packetized stream and thus the common errors for the Internet video are due to packet loss, caused by buffer overflows in routers, late arrival of packets, and bit errors in the network. This loss results in single or multiple macroblock losses in the decoding process and causes severe degradation in perceived quality and error propagation. We present a perceptual preprocessor based on the insensitivity of the human visual system to the mild changes in pixel intensity in order to segment video into regions according to perceptibility of picture changes. With the information of segmentation, we determine which macroblocks require motion estimation and then which macroblocks need to be included in the second layer. The second layer contains the coarse (less quantized) version of the most perceptually-critical picture information to provide redundancy used to reconstruct lost coding blocks. This information is transmitted in a separate packet, which provides path and time diversities when packet losses are uncorrelated. This combination of methods provides a significant improvement in received quality when losses occur, without significantly degrading the video in a low-bit-rate video channel. Our proposed scheme is easily scalable to various data bitrates, picture quality, and computational complexity for use on different platforms. Because the data in our layered video stream is standards-compliant, our proposed schemes require no extra non-standard device to encode/decode the video and they are easily integrated into the current video standards such as H.261/263, MPEG1/MPEG2 and the forthcoming MPEG4.
{"title":"A perceptual-based video coder for error resilience","authors":"Yi-jen Chiu","doi":"10.1109/DCC.1999.785678","DOIUrl":"https://doi.org/10.1109/DCC.1999.785678","url":null,"abstract":"Summary form only given. Error resilience is an important requirement when errors occur during video transmission. The video transmitted over the Internet is usually a packetized stream and thus the common errors for the Internet video are due to packet loss, caused by buffer overflows in routers, late arrival of packets, and bit errors in the network. This loss results in single or multiple macroblock losses in the decoding process and causes severe degradation in perceived quality and error propagation. We present a perceptual preprocessor based on the insensitivity of the human visual system to the mild changes in pixel intensity in order to segment video into regions according to perceptibility of picture changes. With the information of segmentation, we determine which macroblocks require motion estimation and then which macroblocks need to be included in the second layer. The second layer contains the coarse (less quantized) version of the most perceptually-critical picture information to provide redundancy used to reconstruct lost coding blocks. This information is transmitted in a separate packet, which provides path and time diversities when packet losses are uncorrelated. This combination of methods provides a significant improvement in received quality when losses occur, without significantly degrading the video in a low-bit-rate video channel. Our proposed scheme is easily scalable to various data bitrates, picture quality, and computational complexity for use on different platforms. Because the data in our layered video stream is standards-compliant, our proposed schemes require no extra non-standard device to encode/decode the video and they are easily integrated into the current video standards such as H.261/263, MPEG1/MPEG2 and the forthcoming MPEG4.","PeriodicalId":103598,"journal":{"name":"Proceedings DCC'99 Data Compression Conference (Cat. No. PR00096)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127526337","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 mini-corpus of twelve 'calibrated' binary-data files have been produced for systematic evaluation of compression algorithms. These are generated within the framework of a deterministic theory of string complexity. Here the T-complexity of a string x (measured in taugs) is defined as C/sub T/(x/sub i/)=/spl Sigma//sub i/log/sub 2/(k/sub i/+1), where the positive integers k/sub i/ are the T-expansion parameters for the corresponding string production process. C/sub T/(x) is observed to be the logarithmic integral of the total information content I/sub x/ of x (measured in nats), i.e., C/sub T/(x)=li(I/sub x/). The average entropy is H~/sub x/=I/sub x//|x|, i.e., the total information content divided by the length of x. Thus C/sub T/(x)=li(H~/sub x//spl times/|x|). Alternatively, the information rate along a string may be described by an entropy function H/sub x/(n),0/spl les/n/spl les/|x| for the string. Assuming that H/sub x/(n) is continuously integrable along the length of the x, then I/sub x/=/spl int//sub 0//sup |/x|H/sub x/(n)/spl delta/n. Thus C/sub T/(x)=li(/spl int//sub 0//sup |/x|H/sub x/(n)/spl delta/n). Solving for H/sub x/(n): that is differentiating both sides and rearranging, we get: H/sub x/(n)=(/spl delta/C/sub T/(x|n)//spl delta/n)/spl times/log/sub e/(li/sup -1/(C/sub T/(x|/sub n/))). With x being in fact discrete, and the T-complexity function being computed in terms of the discrete T-augmentation steps, we may accordingly re-express the equation in terms of the T-prefix increments: /spl delta/n/spl ap//spl Delta//sub i/|x|=k/sub i/|p/sub i/|; and from the definition of C/sub T/(x): /spl delta/C/sub T/(x) is replaced by /spl Delta//sub i/C/sub T/(x)=log/sub 2/(k/sub i/+1). The average slope over the i-th T-prefix p/sub i/ increment is then simply (/spl Delta//sub i/C/sub T/(x))/(/spl Delta//sub i/|x|)=(log/sub 2/(k/sub i/+1))/(k/sub i/|p/sub i/|). The entropy function is now replaced by a discrete approximation.
只提供摘要形式。12个“校准”二进制数据文件的迷你语料库已经产生了压缩算法的系统评估。这些都是在弦复杂性的确定性理论框架内生成的。这里，字符串x的T-复杂度(以标签为单位)定义为C/下标T/(x/下标i/)=/spl Sigma//下标i/log/下标2/(k/下标i/+1)，其中正整数k/下标i/是对应的字符串生产过程的T-展开参数。观察到C/ T/(x)是x(以纳特为单位)的总信息量I/下标x/的对数积分，即C/ T/(x)=li(I/下标x/)。平均熵为H~/sub x/=I/sub x//|x|，即总信息量除以x的长度，因此C/sub T/(x)=li(H~/sub x//spl乘以/|x|)。或者，沿着字符串的信息速率可以用熵函数H/sub x/(n)来描述，对于字符串，0/spl les/n/spl les/|x|。假设H/下标x/(n)沿x的长度连续可积，则I/下标x/=/spl int//下标0//sup |/x|H/下标x/(n)/spl /n。因此C / sub T /李(x) = (spl int / / sub x | 0 | / /晚餐/ H / sub x / (n) / splδ/ n)。求解H/下标x/(n)也就是两边求导并重新排列，我们得到H/下标x/(n)=(/spl /C/ T/(x|n)//spl /(n) /spl乘以/log/ e/(li/sup -1/(C/下标T/(x|/下标n/)))由于x实际上是离散的，并且t -复杂度函数是用离散的t增积步骤来计算的，因此我们可以用t前缀增量来重新表示方程:/spl delta/n/spl ap//spl delta/ /下标i/|x|=k/下标i/|p/下标i/|;由C/ T/(x)的定义:/spl /C/ T/(x)被/spl //下标i/C/下标T/(x)=log/下标2/(k/下标i/+1)所取代。第i个T前缀p/下标i/增量的平均斜率为(/spl Delta//下标i/C/下标T/(x))/(/spl Delta//下标i/|x|)=(log/下标2/(k/下标i/+1) /(k/下标i/|p/下标i/|))。熵函数现在被一个离散的近似代替了。
{"title":"Towards a calibrated corpus for compression testing","authors":"M. Titchener, P. Fenwick, M. C. Chen","doi":"10.1109/DCC.1999.785711","DOIUrl":"https://doi.org/10.1109/DCC.1999.785711","url":null,"abstract":"Summary form only given. A mini-corpus of twelve 'calibrated' binary-data files have been produced for systematic evaluation of compression algorithms. These are generated within the framework of a deterministic theory of string complexity. Here the T-complexity of a string x (measured in taugs) is defined as C/sub T/(x/sub i/)=/spl Sigma//sub i/log/sub 2/(k/sub i/+1), where the positive integers k/sub i/ are the T-expansion parameters for the corresponding string production process. C/sub T/(x) is observed to be the logarithmic integral of the total information content I/sub x/ of x (measured in nats), i.e., C/sub T/(x)=li(I/sub x/). The average entropy is H~/sub x/=I/sub x//|x|, i.e., the total information content divided by the length of x. Thus C/sub T/(x)=li(H~/sub x//spl times/|x|). Alternatively, the information rate along a string may be described by an entropy function H/sub x/(n),0/spl les/n/spl les/|x| for the string. Assuming that H/sub x/(n) is continuously integrable along the length of the x, then I/sub x/=/spl int//sub 0//sup |/x|H/sub x/(n)/spl delta/n. Thus C/sub T/(x)=li(/spl int//sub 0//sup |/x|H/sub x/(n)/spl delta/n). Solving for H/sub x/(n): that is differentiating both sides and rearranging, we get: H/sub x/(n)=(/spl delta/C/sub T/(x|n)//spl delta/n)/spl times/log/sub e/(li/sup -1/(C/sub T/(x|/sub n/))). With x being in fact discrete, and the T-complexity function being computed in terms of the discrete T-augmentation steps, we may accordingly re-express the equation in terms of the T-prefix increments: /spl delta/n/spl ap//spl Delta//sub i/|x|=k/sub i/|p/sub i/|; and from the definition of C/sub T/(x): /spl delta/C/sub T/(x) is replaced by /spl Delta//sub i/C/sub T/(x)=log/sub 2/(k/sub i/+1). The average slope over the i-th T-prefix p/sub i/ increment is then simply (/spl Delta//sub i/C/sub T/(x))/(/spl Delta//sub i/|x|)=(log/sub 2/(k/sub i/+1))/(k/sub i/|p/sub i/|). The entropy function is now replaced by a discrete approximation.","PeriodicalId":103598,"journal":{"name":"Proceedings DCC'99 Data Compression Conference (Cat. No. PR00096)","volume":"48 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125258810","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}
Recent rate-distortion analyses of image transform coders are based on a trade-off between the lossless coding of coefficient positions versus the lossy coding of the coefficient values. We propose spike processes as a tool that allows a more fundamental trade-off, namely between lossy position coding and lossy value coding. We investigate the Hamming distortion case and give analytic results for single and multiple spikes. We then consider upper bounds for a single Gaussian spike with squared error distortion. The obtained results show a rate distortion behavior which switches from linear at low rates to exponential at high rates.
{"title":"Rate-distortion analysis of spike processes","authors":"C. Weidmann, M. Vetterli","doi":"10.1109/DCC.1999.755657","DOIUrl":"https://doi.org/10.1109/DCC.1999.755657","url":null,"abstract":"Recent rate-distortion analyses of image transform coders are based on a trade-off between the lossless coding of coefficient positions versus the lossy coding of the coefficient values. We propose spike processes as a tool that allows a more fundamental trade-off, namely between lossy position coding and lossy value coding. We investigate the Hamming distortion case and give analytic results for single and multiple spikes. We then consider upper bounds for a single Gaussian spike with squared error distortion. The obtained results show a rate distortion behavior which switches from linear at low rates to exponential at high rates.","PeriodicalId":103598,"journal":{"name":"Proceedings DCC'99 Data Compression Conference (Cat. No. PR00096)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125838046","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}
This paper shows the existence of binary pseudowavelets, bases on the binary domain that exhibit some of the properties of wavelets, such as multiresolution reconstruction and compact support. The binary pseudowavelets are defined on B/sup n/ (binary vectors of length n) and are operated upon with the binary operators logical and, and exclusive or. The forward transform, or analysis, is the decomposition of a binary vector into its constituent binary pseudowavelets. Binary pseudowavelets allow multiresolution, progressive reconstruction of binary vectors by using progressively more coefficients in the inverse transform. Binary pseudowavelets bases, being sparse matrices, also provide for fast transforms; moreover pseudowavelets rely on hardware-friendly operations for efficient software and hardware implementation.
{"title":"Binary pseudowavelets and applications to bilevel image processing","authors":"S. Pigeon, Yoshua Bengio","doi":"10.1109/DCC.1999.755686","DOIUrl":"https://doi.org/10.1109/DCC.1999.755686","url":null,"abstract":"This paper shows the existence of binary pseudowavelets, bases on the binary domain that exhibit some of the properties of wavelets, such as multiresolution reconstruction and compact support. The binary pseudowavelets are defined on B/sup n/ (binary vectors of length n) and are operated upon with the binary operators logical and, and exclusive or. The forward transform, or analysis, is the decomposition of a binary vector into its constituent binary pseudowavelets. Binary pseudowavelets allow multiresolution, progressive reconstruction of binary vectors by using progressively more coefficients in the inverse transform. Binary pseudowavelets bases, being sparse matrices, also provide for fast transforms; moreover pseudowavelets rely on hardware-friendly operations for efficient software and hardware implementation.","PeriodicalId":103598,"journal":{"name":"Proceedings DCC'99 Data Compression Conference (Cat. No. PR00096)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129891432","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}
It is well known that Shannon's separation result does not hold under finite computation or finite delay constraints, thus joint source-channel coding is of great interest for practical reasons. For progressive source-channel coding systems, efficient codes have been proposed for feedforward channels and the important problem of rate allocation between the source and channel codes has been solved. For memoryless channels with feedback, the rate allocation problem was studied by Chande et al. (1998). In this paper, we consider the case of fading channels with feedback. Feedback routes are provided in many existing standard wireless channels, making rate allocation with feedback a problem of considerable practical importance. We address the question of rate allocation between the source and channel codes in the forward channel, in the presence of feedback information and under a distortion cost function. We show that the presence of feedback shifts the optimal rate allocation point, resulting in higher rates for error-correcting codes and smaller overall distortion. Simulations on both memoryless and fading channels show that the presence of feedback allows up to 1 dB improvement in PSNR compared to the similarly optimized feedforward scheme.
{"title":"Progressive joint source-channel coding in feedback channels","authors":"Jin Lu, Aria Nosratinia, B. Aazhang","doi":"10.1109/DCC.1999.755663","DOIUrl":"https://doi.org/10.1109/DCC.1999.755663","url":null,"abstract":"It is well known that Shannon's separation result does not hold under finite computation or finite delay constraints, thus joint source-channel coding is of great interest for practical reasons. For progressive source-channel coding systems, efficient codes have been proposed for feedforward channels and the important problem of rate allocation between the source and channel codes has been solved. For memoryless channels with feedback, the rate allocation problem was studied by Chande et al. (1998). In this paper, we consider the case of fading channels with feedback. Feedback routes are provided in many existing standard wireless channels, making rate allocation with feedback a problem of considerable practical importance. We address the question of rate allocation between the source and channel codes in the forward channel, in the presence of feedback information and under a distortion cost function. We show that the presence of feedback shifts the optimal rate allocation point, resulting in higher rates for error-correcting codes and smaller overall distortion. Simulations on both memoryless and fading channels show that the presence of feedback allows up to 1 dB improvement in PSNR compared to the similarly optimized feedforward scheme.","PeriodicalId":103598,"journal":{"name":"Proceedings DCC'99 Data Compression Conference (Cat. No. PR00096)","volume":"159-160 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122126653","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}
This paper introduces a new method for reducing the number of distance calculations in the generalized Lloyd algorithm (GLA), which is a widely used method to construct a codebook in vector quantization. The reduced comparison search detects the activity of the code vectors and utilizes it on the classification of the training vectors. For training vectors whose current code vector has not been modified, we calculate distances only to the active code vectors. A large proportion of the distance calculations can be omitted without sacrificing the optimality of the partition. The new method is included in several fast GLA variants reducing their running times over 50% on average.
{"title":"Reduced comparison search for the exact GLA","authors":"T. Kaukoranta, P. Fränti, O. Nevalainen","doi":"10.1109/DCC.1999.755651","DOIUrl":"https://doi.org/10.1109/DCC.1999.755651","url":null,"abstract":"This paper introduces a new method for reducing the number of distance calculations in the generalized Lloyd algorithm (GLA), which is a widely used method to construct a codebook in vector quantization. The reduced comparison search detects the activity of the code vectors and utilizes it on the classification of the training vectors. For training vectors whose current code vector has not been modified, we calculate distances only to the active code vectors. A large proportion of the distance calculations can be omitted without sacrificing the optimality of the partition. The new method is included in several fast GLA variants reducing their running times over 50% on average.","PeriodicalId":103598,"journal":{"name":"Proceedings DCC'99 Data Compression Conference (Cat. No. PR00096)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130475499","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 minimum redundancy prefix code problem is to determine, for a given list W=[w/sub 1/,...,w/sub n/] of n positive symbol weights, a list L=[l/sub 1/,...,l/sub n/] of n corresponding integer codeword lengths such that /spl Sigma//sub i=1//sup n/2/sup -li//spl les/1 and /spl Sigma//sub i=1//sup n/w/sub i/l/sub i/ is minimized. Let us consider the case where W is already sorted. In this case, the output list L can be represented by a list M=[m/sub 1/,...,m/sub H/], where m(l/sub 1/), for l=1,...,H, denotes the multiplicity of the codeword length l in L and H is the length of the greatest codeword. Fortunately, H is proved to be O(min{log(1/(p/sub 1/)),n}), where p/sub 1/ is the smallest symbol probability, given by w/sub 1///spl Sigma//sub i=1//sup n/w/sub i/. We present the F-LazyHuff and the E-LazyHuff algorithms. F-LazyHuff runs in O(n) time but requires O(min{H/sup 2/,n}) additional space. On the other hand, E-LazyHuff runs in O(nlog(n/H)) time, requiring only O(H) additional space. Finally, since our two algorithms have the advantage of not writing at the input buffer during the code calculation, we discuss some applications where this feature is very useful.
{"title":"Two space-economical algorithms for calculating minimum redundancy prefix codes","authors":"R. Milidiú, A. Pessoa, E. Laber","doi":"10.1109/DCC.1999.755676","DOIUrl":"https://doi.org/10.1109/DCC.1999.755676","url":null,"abstract":"The minimum redundancy prefix code problem is to determine, for a given list W=[w/sub 1/,...,w/sub n/] of n positive symbol weights, a list L=[l/sub 1/,...,l/sub n/] of n corresponding integer codeword lengths such that /spl Sigma//sub i=1//sup n/2/sup -li//spl les/1 and /spl Sigma//sub i=1//sup n/w/sub i/l/sub i/ is minimized. Let us consider the case where W is already sorted. In this case, the output list L can be represented by a list M=[m/sub 1/,...,m/sub H/], where m(l/sub 1/), for l=1,...,H, denotes the multiplicity of the codeword length l in L and H is the length of the greatest codeword. Fortunately, H is proved to be O(min{log(1/(p/sub 1/)),n}), where p/sub 1/ is the smallest symbol probability, given by w/sub 1///spl Sigma//sub i=1//sup n/w/sub i/. We present the F-LazyHuff and the E-LazyHuff algorithms. F-LazyHuff runs in O(n) time but requires O(min{H/sup 2/,n}) additional space. On the other hand, E-LazyHuff runs in O(nlog(n/H)) time, requiring only O(H) additional space. Finally, since our two algorithms have the advantage of not writing at the input buffer during the code calculation, we discuss some applications where this feature is very useful.","PeriodicalId":103598,"journal":{"name":"Proceedings DCC'99 Data Compression Conference (Cat. No. PR00096)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131888574","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 report on the performance evaluation of greedy parsing with a single-step lookahead, denoted as flexible parsing. We also introduce a new fingerprint-based data structure which enables efficient linear-time implementation.
{"title":"The effect of flexible parsing for dynamic dictionary-based data compression","authors":"Yossi Matias, N. Rajpoot, S. C. Sahinalp","doi":"10.1109/DCC.1999.755673","DOIUrl":"https://doi.org/10.1109/DCC.1999.755673","url":null,"abstract":"We report on the performance evaluation of greedy parsing with a single-step lookahead, denoted as flexible parsing. We also introduce a new fingerprint-based data structure which enables efficient linear-time implementation.","PeriodicalId":103598,"journal":{"name":"Proceedings DCC'99 Data Compression Conference (Cat. No. PR00096)","volume":"80 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126037652","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. In sequential lossless data compression algorithms the data stream is often transformed into short subsequences that are modeled as memoryless. Then it is desirable to use any information that each sequence might provide about the behaviour of other sequences that can be expected to have similar properties. Here we examine one such situation, as follows. We want to encode, using arithmetic coding with a sequential estimator, an M-ary memoryless source with unknown parameters /spl theta/, from which we have encoded already a sequence x/sup n/. In addition, both the encoder and the decoder have observed a sequence y/sup n/ that is generated independently by another source with unknown parameters /spl theta//spl tilde/ that are known to be "similar" to /spl theta/ by a pseudodistance /spl delta/(/spl theta/,/spl theta//spl tilde/) that is approximately equal to the relative entropy. Known to both sides is also a number d such that /spl delta/(/spl theta/,/spl theta//spl tilde/)/spl les/d. For a stand-alone memoryless source, the worst-case average redundancy of the (n+1)-th encoding is lower bounded by 0.5(M-1)/n+O(1/n/sup 2/), and the Dirichlet estimator is close to optimal for this case. We show that this bound holds also for the case with side information as described above, meaning that we can improve, at best, the O(1/n/sup 2/)-term. We define a frequency weighted estimator for this. Application of the frequency weighted estimator to to the PPM algorithm (Bell et al., 1989) by weighting order-4 statistics into an order-5 model, with d estimated during encoding, yields improvements that are consistent with the bounds above, which means that in practice we improve the performance by about 0.5 bits per active state of the model, making a gain of approximately 20000 bits on the Calgary Corpus.
{"title":"On taking advantage of similarities between parameters in lossless sequential coding","authors":"J. Åberg","doi":"10.1109/DCC.1999.785670","DOIUrl":"https://doi.org/10.1109/DCC.1999.785670","url":null,"abstract":"Summary form only given. In sequential lossless data compression algorithms the data stream is often transformed into short subsequences that are modeled as memoryless. Then it is desirable to use any information that each sequence might provide about the behaviour of other sequences that can be expected to have similar properties. Here we examine one such situation, as follows. We want to encode, using arithmetic coding with a sequential estimator, an M-ary memoryless source with unknown parameters /spl theta/, from which we have encoded already a sequence x/sup n/. In addition, both the encoder and the decoder have observed a sequence y/sup n/ that is generated independently by another source with unknown parameters /spl theta//spl tilde/ that are known to be \"similar\" to /spl theta/ by a pseudodistance /spl delta/(/spl theta/,/spl theta//spl tilde/) that is approximately equal to the relative entropy. Known to both sides is also a number d such that /spl delta/(/spl theta/,/spl theta//spl tilde/)/spl les/d. For a stand-alone memoryless source, the worst-case average redundancy of the (n+1)-th encoding is lower bounded by 0.5(M-1)/n+O(1/n/sup 2/), and the Dirichlet estimator is close to optimal for this case. We show that this bound holds also for the case with side information as described above, meaning that we can improve, at best, the O(1/n/sup 2/)-term. We define a frequency weighted estimator for this. Application of the frequency weighted estimator to to the PPM algorithm (Bell et al., 1989) by weighting order-4 statistics into an order-5 model, with d estimated during encoding, yields improvements that are consistent with the bounds above, which means that in practice we improve the performance by about 0.5 bits per active state of the model, making a gain of approximately 20000 bits on the Calgary Corpus.","PeriodicalId":103598,"journal":{"name":"Proceedings DCC'99 Data Compression Conference (Cat. No. PR00096)","volume":"51 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114462094","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. We present an edge-preserving image compression technique based on the wavelet transform and iterative constrained least square regularization. This approach treats image reconstruction from lossy image compression as the process of image restoration. It utilizes the edge information detected from the source image as a priori knowledge for the subsequent reconstruction. Image restoration refers to the problem of estimating the source image from its degraded version. The reconstruction of DWT-coded images is formulated as a regularized image recovery problem and makes use of the edge information as the a priori knowledge about the source image to recover the details, as well as to reduce the ringing artifact of the DWT-coded image. To compromise the rate of edge information and DWT-coded image data, a scheme based on generalized finite automata (GFA) is used. GFA is used instead of vector quantization in order to achieve adaptive encoding of the edge image.
{"title":"Finite automata and regularized edge-preserving wavelet transform scheme","authors":"Sung-Wai Hong, P. Bao","doi":"10.1109/DCC.1999.785687","DOIUrl":"https://doi.org/10.1109/DCC.1999.785687","url":null,"abstract":"Summary form only given. We present an edge-preserving image compression technique based on the wavelet transform and iterative constrained least square regularization. This approach treats image reconstruction from lossy image compression as the process of image restoration. It utilizes the edge information detected from the source image as a priori knowledge for the subsequent reconstruction. Image restoration refers to the problem of estimating the source image from its degraded version. The reconstruction of DWT-coded images is formulated as a regularized image recovery problem and makes use of the edge information as the a priori knowledge about the source image to recover the details, as well as to reduce the ringing artifact of the DWT-coded image. To compromise the rate of edge information and DWT-coded image data, a scheme based on generalized finite automata (GFA) is used. GFA is used instead of vector quantization in order to achieve adaptive encoding of the edge image.","PeriodicalId":103598,"journal":{"name":"Proceedings DCC'99 Data Compression Conference (Cat. No. PR00096)","volume":"69 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124225516","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：