R.L. Joshi, T.R. Fischer, R.H. Bamberger, Optimum classification in subband coding of images, in: Proc. of IEEE Internat. Conf. Image Processing, Vol. 2, November 1994, pp. 883--887.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....procedures. If one does not have access to the best CPU s and wishes to achieve the fastest execution, one could opt to omit arithmetic coding and suffer little consequence in PSNR degradation. Intermediary results can be obtained with, for example, Huffman entropy coding. A recentwork [12], which reports similar performance to our arithmetic coded ones at higher rates, uses arithmetic and trellis coded quantization (ACTCQ) with classification in wavelet subbands. However, at rates below about 0.5 bpp, ACTCQ is not as efficient and classification overhead is not insignificant. Note ....

R.L. Joshi, T.R. Fischer, R.H. Bamberger, "Optimum classification in subband coding of images," Proc. 1994.

....most adaptive quantization schemes used for images select the quantizers based on the original data. Thus, these adaptive quantizers require overhead information to be sent to the decoder, whereas adaptive arithmetic coders do not require any overhead information. As an exam ple, both in [5] and [7], from a discrete set of available quantizers the encoder can select the best quantizer for a given block or region of the image. Our goal is to demonstrate that backward adaptive quantization schemes are also useful in image compression. As a starting point we consider the adaptive quantizer ....

....Note that our goal is to demonstrate the potential gain in introducing these techniques. More sophisticated quantizers and schemes taking advantage of interband correlation can be incoporated into our proposed scheme and would give better results. 2. CLASSIFICATION 50 lOO 15o 200 In [7] a method is proposed to classify blocks of sub band coefficients into classes based on their activity . Each class is modelled as a Laplacian source with a different parameter and each block within a subband is thus assigned to a class, and its corresponding Laplaclan model, and is quantized ....

[Article contains additional citation context not shown here]

R. Joshi, T. Fischer, and R. Bamberger, "Optimum classification in subband coding of images," in Proc. of ICIP'9J, vol. 2, (Austin, Texas), pp. 883 887, Oct. 1994.

....Recent advances in wavelet based image coding reveal that this assumption is far from accurate, as testified by the success of coders based on data structures that exploit the dependencies among the wavelet coefficients (e.g. based on zerotrees, morphology, classification based approaches etc. [29, 30, 31]) It can be noted that the methodology advocated in this work can be appropriately modified to include such dependencies, at least partially. For example, using a zerotree data structure [29, 32] it is possible to reduce the number of coefficients needing to be transmitted by grouping sets of ....

R. L. Joshi, T. R. Fischer, and R. H. Bamberger, "Optimum classification in subband coding of images," in Proc. IEEE Int. Conf. Image Proc., (Austin, TX), pp. 883--887, Nov. 1994.

....Recent advances in wavelet based image coding reveal that this assumption is far from accurate, as testified by the success of coders based on data structures that exploit the dependencies among the wavelet coefficients (e.g. based on zerotrees, morphology, classification based approaches etc. [29, 30, 31]) It can be noted that the methodology advocated in this work can be appropriately modified to include such dependencies, at least partially. For example, using a zerotree datastructure [29, 32] it is possible to reduce the number of coefficients needing to be transmitted by grouping sets of ....

R. L. Joshi, T. R. Fischer, and R. H. Bamberger, "Optimum classification in subband coding of images," in Proc. IEEE Int. Conf. Image Proc., (Austin, TX), pp. 883--887, Nov. 1994.

....in the message passing procedure of the sum product algorithm. observations are i.i.d. Gaussian random variables with mean and variance indexed by the state. This model is motivated by image transform coding, where similar 2 D models have been proposed for transform coefficients (e.g. schemes in [9, 10] have the same spirit, though not being exactly equivalent to our model here) that acknowledge the fact of remaining localized dependencies even after decorrelating transform. State 0 State 1 N( 0 ,s 0 ) N( 1 ,s 1 ) P(1 0) P(0 1) P(0 0) P(1 1) Figure 6. Hidden Markov source model with ....

R. L. Joshi, T. R. Fischer, and R. H. Bamberger, "Optimum classification in subband coding of images," in Proc. IEEE Int. Conf. Image Proc., (Austin, TX), pp. 883--887, Nov. 1994. 5 A framework for joint source-channel coding using factor graphs. I. Kozintsev R. Koetter and K. Ramchandran.

....procedures. If one does not have access to the best CPU s and wishes to achieve the fastest execution, one could opt to omit arithmetic coding and suffer little consequence in PSNR degradation. Intermediary results can be obtained with, for example, Huffman entropy coding. A recent work [12], which reports similar performance to our arithmetic coded ones at higher rates, uses arithmetic and trellis coded quantization (ACTCQ) with classification in wavelet subbands. However, at rates below about 0.5 bpp, ACTCQ is not as efficient and classification overhead is not insignificant. Note ....

R.L. Joshi, T.R. Fischer, R.H. Bamberger, "Optimum classification in subband coding of images," Proc. 1994 IEEE Int. Conf. on Image Processing, vol. II, pp. 883--887, Austin, TX, Nov. 1994.

....are forward adaptive. That is, the encoder adaptively assigns the best quantizer from a finite collection of available quantizers to each of the subband coefficient sets by examining in advance the current data set to quantize. Note that forward adaptive quantizers, including the ones in [7] and [11], require overhead information to be sent to the decoder, thus increasing the required overall rate. Alternatively, backward adaptation operates by adjusting the system through prediction of the current data characteristics, where the prediction is based on previously processed data. A good ....

....it is useful to separate the coefficients in a given subband into different classes so that different quantization determined by bit allocation can be applied to each class. This fact is well known and a good summary of the previous classification efforts for image compression can be found in [11] and [13] which also demonstrate successful applications of classification in subband image coding. As mentioned above most previous classification attempts, including [11] and [13] can be categorized under forward adaptation: as a preprocessing stage to quantization, classification is performed ....

[Article contains additional citation context not shown here]

R. Joshi, T. Fischer, and R. Bamberger, "Optimum classification in subband coding of images," in Proc. of ICIP'94, vol. 2, (Austin, Texas), pp. 883--887, Oct. 1994.

....and informatics, N 7034 Trondheim, Norway. positions of the non zero amplitudes (referred to as the significance map) are efficiently coded by means of a zerotree constructed from a hierarchical subband wavelet decomposition. The second category of coding schemes uses block wise classification [3]. For these coding schemes, the transformed image is divided into blocks, and each block is coded with a certain number of bits according to its significance. Typically, the classification is performed by distributing the available bits among the blocks according to their mean squared value. It is ....

....table can be used to code each integer number individually. This approach would improve the performance compared to the fixed length code, and can be improved further by designing variable length codes based on the conditional entropy between neighbor entries in the classification table [3]. In this work, we propose a new strategy for coding of the classification table referred to as hierarchical enumeration. Let the classification table to be coded be described by the integer valued vector n = fn0 ; n1 ; nL Gamma1 g; n i 2 f0; J Gamma 1g: 1) Assume that the sum K = L ....

R. L. Joshi, T. R. Fischer, and R. H. Bamberger, "Optimum classification in subband coding of images," in Proc. Int. Conf. on Image Processing (ICIP), vol. II, pp. 883 -- 886, 1994.

....forward adaptive. That is, the encoder adaptively assigns the best quantizer from a finite collection of available quantizers to each of the subband coefficient sets by examining in advance the current data set to quantize. Note that forward adaptive quantizers, including the ones in [6] and [10], require overhead information to be sent to the decoder, thus increasing the required overall rate. Alternatively, backward adaptation operates by adjusting the system through prediction of the current data characteristics, where the prediction is based on previously processed data. A good ....

....it is useful to separate the coefficients in a given subband into different classes so that different quantization determined by bit allocation can be applied to each class. This fact is well known and a good summary on the previous classification efforts for image compression can be found in [10] and [12] which also demonstrate very successful applications of classification in subband image coding. We note that most previous classification attempts, including [10] and [12] can be categorized under forward adaptation: as a preprocessing stage to quantization, classification is performed ....

[Article contains additional citation context not shown here]

R. Joshi, T. Fischer, and R. Bamberger, "Optimum classification in subband coding of images," in Proc. of ICIP'94, vol. 2, (Austin, Texas), pp. 883--887, Oct. 1994.

....and ensuring that zero can be used as an output codeword with high probability. The codebooks are designed using the algorithm and encoding rule from [85] 9. 2 TCQ subband coders In a remarkable coincidence, at the 1994 International Conference in Image Processing, three research groups [86, 87, 88] presented similar but independently developed image coding algorithms. The main ingredients of the three methods are subband decomposition, classification and optimal rate allocation to di#erent subsets of subband data, and entropy constrained TCQ. These works have been brought together in [68] ....

R. L. Joshi, T. Fischer, and R. H. Bamberger, "Optimum classification in subband coding of images," in Proc. IEEE Int. Conf. Image Proc. (ICIP), vol. 2, (Austin, TX), pp. 883--887, November 1994.

....their ac energy. Kim and Lee[2] have the DCT transformed input block classified according to the perceptual feature. By using their approach, an efficient edge oriented classifier employing the DCT coefficient as feature for classification is used to maintain edge integrity. Recently, Joshi et al.[3] presented an algorithm by maximizing the coding gain for a fixed number of classes. Although these algorithms improve the coding performance as compared to traditional schemes, the major drawback of these algorithm is the high complexity of the classification map, and consequently requires ....

Rajan L. Joshi, Thomas R. Fischer, and Roberto H. Bamberger "Optimum Classification in Subband Coding of Images", Proc. of ICIP, pp883-887, 1994.

....their ac energy. Kim and Lee[2] have the DCTtransformed input block classified according to the perceptual feature. By using their approach, an efficient edge oriented classifier employing the DCT coefficient as feature for classification is used to maintain edge integrity. Recently, Joshi et al.[3] presented an algorithm by maximizing the coding gain for a fixed number of classes. Although these algorithms improve the coding performance as compared to traditional schemes, the major drawback of these algorithm is the high complexity of the classification map, and consequently requires ....

Rajan L. Joshi, Thomas R. Fischer, and Roberto H. Bamberger "Optimum Classification in Subband Coding of Images", Proc. of ICIP, pp883-887, 1994.

.... only modest improvement over standard transform and non hierarchical subband algorithms [3] Recently, a new class of wavelet al..gorithms have demonstrated significantly improved performance over the earlier wavelet coders by exploiting the wavelet s space frequency (S F) compaction properties [4, 5, 6]. These new algorithms recognize that the wavelet transform offers two forms of energy compaction that hold potential for coding gain. First, the wavelet is able to compact the energy of smoothly varying regions in an image into a few low frequency coefficients. Second, overlooked by earlier ....

....allocation of bitrate among the regions. Any side information needed for identifying regions and assigning quantizers to regions should also be included in the bitrate allocation optimization. Various approaches to the space frequency coding problem have been proposed and studied in the literature [4, 5, 6]. They all involve some aspects of spatial classification of wavelet coefficients either prior to, or as part of, the quantization process. Classifications are computed at the encoder and transmitted to the decoder as side information. Classifications have the form of bitplane significance maps, ....

R. Joshi, T. R. Fischer, and R. H. Bamberger, "Optimum classification in subband coding of images." Proceedings ICIP-94, November 1994, pp. 883-887.

....subband image coder based on classification are presented. The simulation results demonstrate the value of classification in subband coding. 1 Preface This paper is the result of a collaboration initiated at the 1994 International Conference on Image Processing, where authors of the three papers [13], 10] and [16] recognized that they had independently developed very similar image coding methods and achieved similar performance. The main ingredients of these image coding algorithms are subband decomposition to remove linear dependence in the image data, classification and optimum rate ....

....dependence in the image data, classification and optimum rate allocation to effectively model and allocate encoding rate to different subsets of the subband data, and entropy constrained trellis coded quantization (ECTCQ) 6] 19] to encode the subbands. The particulars of the algorithms in [13], 10] and [16] vary in the choice of subband filters, the method of implementing the ECTCQ, and the approach to classification of subband data. The goal of this collaboration is to provide a thorough examination of the image coding principles underlying [13] 10] and [16] compare the ....

[Article contains additional citation context not shown here]

R. L. Joshi, T. R. Fischer, and R. H. Bamberger, "Optimum classification in subband coding of images," Proceedings, Int. Conf. Image Proc., vol. 2, pp. 883-887, November 1994.

No context found.

R.L. Joshi, T.R. Fischer, R.H. Bamberger, Optimum classification in subband coding of images, in: Proc. of IEEE Internat. Conf. Image Processing, Vol. 2, November 1994, pp. 883--887.

No context found.

R.L. Joshi, T.R. Fischer, and R.H. Bamberger, "Optimum Classification in Subband Coding of Images," Proceedings, IEEE Int. Conf. on Image Proc., Austin, TX, Vol. II, pp. 883--887, Nov. 1994.

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