M. J. Weinberger, G. Seroussi, and G. Sapiro, "The LOCO-I Lossless Image Compression Algorithm: Principles and Standardization into JPEG-LS," IEEE Trans. on Image Processing, vol. 9, no. 8, pp. 1309--1324, 2000.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....bypassed. In the SPIHT method, we rearrange the binDCT coefficients according to the pattern of the wavelet transform coefficients before applying zerotree processing [50] The binDCT based lossless transforms are compared with two advanced context model based prediction methods: the HP LOCO I [51] and CALIC [52] The results are summarized in Table VIII, showing that the overall compression ratio of the binDCT based method is not as good as these methods. However, the proposed binDCT is much simpler, and it provides a unified framework for both lossy and lossless compression. IX. ....

M. J. Weinberger, G. Seroussi, and G. Sapiro, "The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS," IEEE Trans. Image Processing, vol. 9, pp. 1309--1324, Aug. 2000.

....unrefined predictor Eqn. 2 and corresponds to a texture context 3 . Typically, the probability distribution of the prediction error, R , can be approximated fairly well by a Laplacian distribution with zero mean and a small variance which is correlated with the context [6, pp. 33] [7]. Here, we assume that the prediction error distribution is exactly Laplacian. The arguments and the ensuing conclusions and techniques, however, are largely applicable even when the true distributions deviate from this assumption. Fig. 1.a shows a plot of the probability mass function (pmf) S ....

M. Weinberger, G. Seroussi, and G. Sapiro, "The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS," IEEE Trans. on Image Proc., vol. 9, pp. 1309--1324, Aug. 2000.

....decoding step, the pixel is first decoded and the update follows. Typically, the probability distribution of the prediction error, # = s s, can be approximated fairly well by a Laplacian distribution with zero mean and a small variance which is correlated with the context d [15, pp. 33] 16] [17]. In order to make precise statements, for the following discussion, we assume that the prediction error distribution p(# d) is exactly Laplacian with variance # d determined by d. The arguments and the ensuing conclusions and techniques, however, are largely applicable even when the true ....

Weinberger, Seroussi, and Sapiro, "The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS," IEEE Trans. on Image Processing, vol. 9, pp. 1309-- 1324, Aug. 2000.

....of research man months. Table 1. Image compression standards [12 14, 18 25] Official Area Popular name notation of application Technology Official and de facto lossless compression standards JPEG LS ISO IS 14495 Lossless and near lossless Adopted LOCO I algorithm compression of continuous [ 16,42]: Backward adaptive tone images. Relatively new prediction nonlinear prediction standard. with correction Golomb Rice coding or run legth coding. 11,12,15,20] PNG Newer graphic file format. Prediction dictionary coding Huffman coding. GI1 e Graphic file format widly Dictionary coding ....

Weinberger M.J., Seroussi G., Sapiro G.: The LOCO-I lossless image compression algorithm: principles and standardization into JPEG-LS, IEEE Trans. Image Proc., vol. 9, 2000, pp. 1309-1324.

....when j u i 1 U i 1 j is small. Then if j i = 0, it is highly likely that J i = 0 and, conversely if j i 6= 0, it is highly likely that J i 6= 0. On the contrary when j u i 1 U i 1 j is large, then it is highly likely that J i 6= 0. Using this information we can define two different contexts [24], 25] for J i U i 1 j T q and j i = 0 ) p(J i = 0) p(J i 6= 0) U i 1 j T q or j i 6= 0 ) p(J i 6= 0) p(J i = 0) This information can be exploited by using a bitmap to indicate the more probable event in each context, i.e. in context (C1) we transmit a 0 if J i = 0 and a 1 ....

G. S. M. Weinberger, G. Seroussi, "The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS," IEEE Trans. Image Processing, vol. 9, pp. 1309--1324, August 2000.

....in the decoding step, the pixel is first decoded and the update follows. Typically, the probability distribution of the prediction error, # = s s, can be approximated fairly well by a Laplacian distribution with zero mean and a small variance which is correlated with the context d [7, pp. 33] [9,10]. In order to make precise statements, for the following discussion, we assume that the prediction error distribution p(# d) is exactly Laplacian with variance # d For the experimental results of Section 3, the quantizer Q( s threshold are 1, 2, 3, 4, 6, 10, 15 In order to avoid ....

M. Weinberger, G. Seroussi, G. Sapiro, The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS, IEEE Trans. on Image Processing 9 (2000) 1309--1324.

....variation, emphasizing the lossless JPEG 2000 case. This work was partially supported by the FCT. 12th June 2002 DRAFT IEEE SIGNAL PROCESSING LETTERS, VOL. 9, NO. 8, AUG. 2002 I. INTRODUCTION It has been shown recently [1] that the performance of lossless image coding methods such as JPEG LS [2, 3], lossless JPEG 2000 [4, 5] or CALIC [6] can be improved by a simple preprocessing technique, which can be applied to images with sparse histograms. The method is illustrated in Fig. 1. Before coding, the image is subject to a mapping T , which packs its histogram. Applying T 1 to the uncoded ....

M. J. Weinberger, G. Seroussi, and G. Sapiro, "The LOCO-I lossless image compression algorithm: principles and standardization into JPEG-LS", IEEE Trans. on Image Processing, vol. 9, no. 8, pp. 1309--1324, Aug. 2000.

....near lossless coding is not satisfactory. In the well known wavelet multiresolution decompositions, the coe cients of the lters are xed. They are not calculated in order to t the image the best. Moreover, most of the lossless image compression algorithms used today, like CALIC [26] and LOCO I [25], are not associated with multi resolution decompositions. On the other hand, they adapt to the image using context based predictors. Bi orthogonal wavelet decompositions are e cient for lossy and near lossless image compression, hence they are used in the ISO JPEG2000 standard. The lifting scheme ....

....both possible. In the next section we present the adapted lifting scheme framework which is shared by both locally and globally adapted estimation methods. In Sections III and IV, the LAE and GAE methods are shown with details. These methods are compared with well known codecs (SPIHT [21] LOCO I [25], CALIC [26] and JASPER [1] in Section V. In the following, Z denotes the set of all integers. For a matrix A, its transpose is written A . Underlined lower case letters denote vectors, which are identi ed with the column matrix of their coordinates. II. Adapted lifting schemes with truncated ....

M. J. Weinberger, G. Seroussi and G. Sapiro, The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS, IEEE Trans. Image Processing, vol 9, pp. 13091324, Aug. 2000.

....) #labelled pixels in the image) we obtain the coding cost in bits pixel. The minimum attainable coding cost is the entropy (in bits pixel) of the generating process. As the computation of P (x ) is intractable in MRF models, we compare the TSBN results to those from the lossless JPEG LS codec [44, 45], available from http: www.hpl.hp.com loco . 12 3.3.1 Coding cost under a TSBN model Using a TSBN to model the distribution of images, the marginal likelihood of a label image x can be calculated eciently at the root node X of the tree (see Appendix A.2) Below we will also consider the ....

M. Weinberger, G. Seroussi, and G. Sapiro. The LOCO-I Lossless Image Compression Algorithm: Principles and Standardization into JPEG-LS, 1999.

....Introduction The rapidly evolving information age, invoked by the colossal developments in telecommunications and in the satellite, cellular, and networking technologies, ensues the pressing need for eciently representing images and video signals. Practical modern compression techniques (cf. 20] [22] and references therein) are observed to be ecient and relatively robust for the storage and transmission of images and video signals. Contrary to the abundance of reported work in the signal , image , and video processing communities (cf. 13] 19] and references therein) the problem of ....

M. Weinberger, G. Seroussi, and G. Sapiro, \The LOCO-I Lossless Image Compression Algorithm: Principles and Standardization into JPEG-LS," Computer Systems Laboratory, HPL-98-197, Nov. 1998.

....consists of scanning the data array, constructing a model of the data, employing a predictor corresponding to the model, and then encoding the di#erence between the prediction and the actual outcome. Examples for predictive coding include LPC based voice coders (e.g. 1] and image coders (e.g. [2]) The compression e#ciency of such schemes naturally boils down to the e#ciency of the prediction scheme employed. Now, assuming the predictive coding part of the scheme is fixed (i.e. the way that the di#erence between the predicted value and the actual outcome is encoded) a degree of freedom ....

....assumed ordered as a one dimensional time series. In such problems sequentiality usually dictates only one possibility for scanning the data, namely, the direction of the flow of time. However, when the dimension of the data array, d, is larger than 1 (e.g. in image and video coding applications [3, 4, 5, 2]) there is no natural direction # Authors are with the Department of Electrical Engineering, Technion Israel Institute of Technology, Haifa 32000, Israel. merhav ee.technion.ac.il, tsachy tx.technion.ac.il. 1 of the flow of time and the question of the optimal scheme for scanning and ....

M. Weinberger, G. Seroussi, and G. Sapiro. The loco-i lossless image compression algorithm: Principles and standardization into jpeg-ls. IEEE Trans. Image Proc., 1998. Submitted. Available as Hewlett-Packard Laboratories Technical Report No. HPL-98-193, 1998.

....DCT coecients. In the SPIHT method, we rearrange the binDCT coecients according to the pattern of the wavelet transform coecients before applying zerotree processing [50] The binDCT based lossless transforms are compared with two advanced context model based prediction methods: the HP LOCO I [51] and CALIC [52] The results are summarized in Table VIII, showing that the overall compression ratio of the binDCT based method is not as good as these methods. However, the proposed binDCT is much simpler, and it provides a uni ed framework for both lossy and lossless compression. IX. ....

M. J. Weinberger, G. Seroussi, and G. Sapiro, \The loco-i lossless image compression algorithm: principles and standardization into jpeg-ls," IEEE Trans. Img. Process., vol. 9 (8), pp. 1309-1324, Aug. 2000.

....annoying in smooth regions when the quantization is coarse. These artifacts are called banding due to their appearance; it looks like bands were introduced to separate regions in the image. The JPEG LS (near lossless still image compression standard) based on the LOCO (Low Complexity) algorithm [34] presents these artifacts. This algorithm is based on quantization of an non linear prediction for each pixel and is described later. The examples in Figs. 2.7, 2.8 and 2.9 illustrate the banding artifacts for the JPEG LS algorithm. 13 Figure 2.7: Lenna Compressed by JPEG LS (LOCO) Figure ....

....based on the quantization step used in the JPEG algorithms. In 27 this way we exploit the characteristics of the compression technique (and therefore indirectly the original image) while avoiding weak assumptions about the distribution. For the JPEG LS using the LOCO (Low Complexity) algorithm [34], an non linear prediction P is made based on previous pixels in a scanning order. The prediction error E, computed as the di erence between the actual pixel value X and the non linear prediction P , is uniformly quantized using a xed quantization step QS. The amount of banding artifacts depends ....

[Article contains additional citation context not shown here]

Marcelo Weinberger, Gadiel Seroussi and Guillermo Shapiro. The LOCO-I Lossless Image Compression Algorithm: Principles and Standardization into JPEG-LS. Hewlett Packard Labs, Technical Reports no. HPL-98-193, 1998.

....been found for the LZ algorithm. The second type of universal code is based on an explicit probability model. One good example is Rissanen s context algorithm, which builds on the finite state machine defined by context trees. It is slower than LZ, but provides better compression for image data. Weinberger, Seroussi and Sapiro s (1998) LOCO algorithm is based on the context modeling idea and has been selected as the new lossless (or nearly lossless) image compression standard JPEG LS 2000. The mean per symbol length of the LZ code approaches the entropy at the rate of O(1= log n) n being the length of the string, no matter ....

Weinberger, M. J., Seroussi, G., and Sapiro, G. (1998). "The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS," IEEE Trans. Image Proc., (submitted). Available as Hewlett-Packard Laboratories Technical Report No. HPL-98-193.

....errors. Already coded errors G( b( S( bits e i;j Fig. 2. Compression with the Golomb based Coder Larger b s are better when S(e i;j ) might be large, since they make codeword lengths grow more slowly as S(e i;j ) increases. b is a kind of uncertainty parameter because of this. JPEG LS [2] also uses Golomb codes and S( 2] credits S( to R.F. Rice. As the appendix discusses, b( and oe( s similarity is not a coincidence. b( is also somewhat similar to the uncertainty parameter estimators in JPEGLS and FELICS [5] One difference is that b( depends on contiguously located ....

....b( S( bits e i;j Fig. 2. Compression with the Golomb based Coder Larger b s are better when S(e i;j ) might be large, since they make codeword lengths grow more slowly as S(e i;j ) increases. b is a kind of uncertainty parameter because of this. JPEG LS [2] also uses Golomb codes and S( [2] credits S( to R.F. Rice. As the appendix discusses, b( and oe( s similarity is not a coincidence. b( is also somewhat similar to the uncertainty parameter estimators in JPEGLS and FELICS [5] One difference is that b( depends on contiguously located nearby errors instead of errors that may ....

[Article contains additional citation context not shown here]

M. Weinberger, G. Seroussi, and G. Sapiro, "The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS," Tech. Rep. HPL-98-193R1, Hewlett-Packard Laboratories, October 1999, www.hpl.hp.com/research/itc/ csl/vcd/infotheory/loco.html.

....originates from Magnetic Resonance Imaging (MRI) and Computerized Tomography (CT) we will focus on these types of images. MRI and CT images are volume images, in which all three dimensions are spatial. There exist many lossless image compressors (LJPEG [1] CALIC [4, 5] BTPC [6, 7] JPEG LS [8, 9], etc. but most of them do not take the third dimension into account. In this case, the frames of an image must be compressed independently of each other. However, it is clear that the third dimension introduces some interframe redundancy, which could be exploited to attain higher compression. ....

M. Weinberger, G. Seroussi, and G. Sapiro, "The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS," Tech. Rep. HPL-98-193, HP Computer Systems Laboratory, Nov. 1998, http://www.hpl.hp.com/techreports/98.

....cropping) x Region of interest coding by progression x Limited memory implementations. The JPEG 2000 project was motivated by Ricoh s submission of the CREW algorithm [1,2] to an earlier standardization effort for lossless and near lossless compression (now known as JPEG LS) Although LOCO I [3] was ultimately selected as the basis for JPEG LS, it was recognized that CREW provided a rich set of features 1 Department of ECE, University of Arizona, Tucson, AZ, marcellin,bilgin ece.arizona.edu 2 Ricoh Silicon Valley, Menlo Park, CA, gormish,boliek rsv.ricoh.com 2 worthy of a new ....

M. Weinberger, G. Seroussi, and G. Sapiro, "The LOCO-I lossless image compression algorithm: principles and standardization into JPEG-LS," submitted to IEEE Trans. on Image Proc.

No context found.

M. J. Weinberger, G. Seroussi, and G. Sapiro, "The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS," Trans. Image Processing, vol. IP-9, pp. 1309--1324, Aug. 2000. Available as Hewlett-Packard Laboratories Technical Report HPL-98-193(R.1).

No context found.

M. J. Weinberger, G. Seroussi, and G. Sapiro, "The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS," 1998. Submitted to IEEE Trans. Image Proc. Available as Hewlett-Packard Laboratories Technical Report No. HPL-98-193.

....x t 1 . In universal coding for a parametric class of sources, the above probability assignment is designed to simultaneously best match every possible source within this class. For example, the context (or nite memory) model [1, 2] has been successfully applied to lossless image compression [3, 4, 5, 6], an application which serves as the main motivation for this work. According to this model, the conditional probability of each symbol, given the entire past, depends only on a bounded, but possibly varying number of the most recent past symbols, referred to as context. In this case, the ....

....zero, henceforth referred to as centered TSGD. According to this distribution, the probability of an integer value x of the prediction error (x = 0; 1; 2; is proportional to jxj , where 2 (0; 1) controls the two sided exponential decay rate. When combined with a context model as in [4, 5], the TSG model is attractive also because there is only one parameter ( per context, although the alphabet is in principle in nite (and in practice nite but quite large, e.g. 8 bits per pixel) This allows for a modeling strategy based on a fairly large number of contexts at a reasonable ....

[Article contains additional citation context not shown here]

M. J. Weinberger, G. Seroussi, and G. Sapiro, \The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS," 1998. Submitted to IEEE Trans. Image Proc. Available as Hewlett-Packard Laboratories Technical Report. 20

....matrix construction it is assumed that cost(i, j) cost(j, i) so that only half of the matrix must be computed. We generated matrices using various cost functions, such the plain entropy of the prediction error, prediction error entropy after application of the Median Edge Detector described in [3], and number of bits generated by an actual compression scheme with the chosen lossless or lossy encoding. Each cost function represents a trade off between the compression achieved and the complexity of computing it. The decorrelation sequence that minimizes the total cost is determined by ....

....of allowing smaller displacements while limiting the complexity of a full search. After each plane has been decorrelated, a standard image compression algorithm can be applied to the planes (if intra coded) or to their prediction error (if inter coded) For our experiments we used JPEG LS [3] for lossless and near lossless compression, and JPEG [4] for lossy compression. Higher compression ratios in lossy mode, at the cost of increased computational complexity, can be obtained with the upcoming JPEG2000 standard. Modulo Reduction When subtracting a plane from another, the alphabet ....

[Article contains additional citation context not shown here]

M. Weinberger, G. Seroussi and G. Sapiro, "The LOCO-I Lossless Image Compression Algorithm: Principles and Standardization into JPEG-LS", IEEE Trans. on Image Processing, Vol. 9, No. 8, August

.... on any integer value, an assumption that, in the context of exponential distributions, is well approximated in practice by the use of large symbol alphabets (e.g. 8 bits per pixel) Although the centered TSGD is an appropriate model for memoryless image compression schemes, it has been observed [3, 4] that prediction errors in context based schemes [5, 6, 3, 4] exhibit a DC offset, and a more appropriate model is given by an off centered TSGD. This model is also useful for better capturing the two adjacent modes often observed in empirical context dependent histograms of prediction errors. ....

.... of exponential distributions, is well approximated in practice by the use of large symbol alphabets (e.g. 8 bits per pixel) Although the centered TSGD is an appropriate model for memoryless image compression schemes, it has been observed [3, 4] that prediction errors in context based schemes [5, 6, 3, 4] exhibit a DC offset, and a more appropriate model is given by an off centered TSGD. This model is also useful for better capturing the two adjacent modes often observed in empirical context dependent histograms of prediction errors. More specifically, in this paper we consider integer ....

[Article contains additional citation context not shown here]

M. J. Weinberger, G. Seroussi, and G. Sapiro, "The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS," 1998. Submitted to IEEE Trans. Image Proc. Available as Hewlett-Packard Laboratories Technical Report No. HPL-98-193.

No context found.

M. J. Weinberger, G. Seroussi, and G. Sapiro, "The LOCO-I Lossless Image Compression Algorithm: Principles and Standardization into JPEG-LS," IEEE Trans. on Image Processing, vol. 9, no. 8, pp. 1309--1324, 2000.

No context found.

WEINBERGER M. J., SEROUSSI G., SAPIRO G.: The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS. IEEE Transactions on Image Processing 9, 8 (Aug. 2000), 1309--1324.

No context found.

M. J. Weinberger, G. Seroussi, and G. Shapiro, "The LOCO-I lossless image compression algorithm: principles and standardization into JPEG-LS," IEEE Trans. Image Processing, Vol. 9, pp. 1309-1324, 2000.

No context found.

M. Weinberger, G. Seroussi, G. Sapiro: "The LOCO-I Lossless Image Compression Algorithm: Principles and Standardization into JPEG-LS", IEEE Trans. on Image Processing, Vol. 9, August 2000, pp.1309-1324.

No context found.

Weinberger, M. J., Seroussi, G., and Sapiro, G. (1998). "The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS," IEEE Trans. Image Proc., (submitted). Available as Hewlett-Packard Laboratories Technical Report No. HPL-98-193.

No context found.

Weinberger, M., Seroussi, G., Sapiro, G.: The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS. IEEE Transactions on Image Processing 9 (2000) 1309--1324

No context found.

M. Weinberger, G. Seroussi, and G. Sapiro, "The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS," IEEE Trans. Image Processing, vol. 9, pp. 1309--1324, Aug. 2000.

No context found.

M Weinberger and G Seroussi. The LOCO-I Lossless Image Compression Algorithm: Principles and Standardization into JPEG-LS. 343, 347

No context found.

M. Weinberger, G. Seroussi, and G. Sapiro, "The LOCO-I Lossless Image Compression Algorithm: Principles and Standardization into JPEG-LS," Computer Systems Laboratory, Hewlett-Packard, Palo Alto, CA, HPL-98-193, Nov. 1998.

No context found.

M.J. Weinberger, G. Seroussi, and G. Sapiro, "The LOCO-I lossless image compression algorithm: principles and standardization into JPEG-LS," IEEE Trans. Image Proc., vol. 9, pp. 1309--1324, Aug. 2000.

No context found.

M. Weinberger, G. Seroussi, G. Sapiro: "The LOCO-I Lossless Image Compression Algorithm: Principles and Standardization into JPEG-LS", IEEE Trans. on Image Processing, Vol. 9, August 2000, pp.1309-1324.

No context found.

M. J. Weinberger, G. Seroussi, and G. Sapiro, "The LOCO-I lossless image compression algorithm: principles and standardization into JPEG-LS," IEEE Trans. on Image Processing, vol. 9, no. 8, pp. 1309-- 1324, Aug. 2000.

No context found.

M. J. Weinberger, G. Seroussi, and G. Sapiro, "The LOCO-I lossless image compression algorithm: principles and standardization into JPEG-LS", IEEE Transactions on Image Processing, 9 (8), pp.1309-24, 2000.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC