R. F. Chang, W. T. Chen, and J. S. Wang, "Image sequence coding using adaptive tree-structured vector quantisation with multipath searching," IEE Proc. I, vol. 139, pp. 9--14, February 1992.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....current input distribution is estimated based on previously quantized samples. In some adaptive quantization techniques[5, 6] the ob jective is to adjust the support region of a uniform scalar quantizcr, so that this quantizcr can bc used in conjunc tion with a predictor in a DPCM system. In [7] an initial trcc structured vector quantizcr (TSVQ) is first designed, with a rate higher than the rate available for transmission. Then the adaptive algorithm chooses which subtree of the previously designed trcc has to bc used at every instant, by keeping counts of the number of samples that ....

R. Chang, W. Chen, and J. Wang, "Image sequence coding using adaptive tree-structured vector quantisation with multipath searching," IEE Proc. I, vol. 139, pp. 9 14, Feb. 1992.

....in searching the tree, PTSVQ permits the user to vary the compression rate. A drawback of PTSVQ is its variable index rate and variable search time. On the positive side, the variable search time conveniently masks the variable execution time of VPC. TSVQ has also been extended to multi path TSVQ [CCW91] which searches for the four best candidates in the tree. Overall, PTSVQ seems the most viable fast search technique. 2.7 Previous Implementations Table 2.1 summarizes previous VQ implementations. Most of the early VQ implementation were for speech compression. Since speech has orders of ....

R. Chang, W. Chen, and J. Wang. Image sequence coding using adaptive treestructured vector quantization with multipath searching. In IEEE Int. Conf. Acous., Speech, and Signal Proc., pages 2281--2284, 1991.

....by theory and that actually obtained in real implementations. Indeed, the nonstationary nature of sources common in practical applications has prompted a search for more general VQ algorithms that are capable of adapting to changing source statistics as coding progresses. Such algorithms (e.g. [2 16]) are called adaptive vector quantization (AVQ) In this paper, we first briefly outline a mathematical definition of AVQ which describes the structure of AVQ Support was provided by an AT T Ph.D. Scholarship This work was performed while the author was with the Department of Electrical ....

....quantizer that achieves this bound as the dimension of the quantizer becomes infinitely large [20] This theoretic asymptotic optimality of VQ has inspired its use in many applications. However, most sources of practical interest are, in reality, nonstationary. A number of AVQ algorithms (e.g. [2 16]) have been introduced to provide more efficient coding in these applications. These AVQ algorithms compensate for the changing source statistics associated with nonstationary sources by periodically updating the VQ codebook. We have developed a mathematical definition to describe the operation ....

[Article contains additional citation context not shown here]

Ruey-Feng Chang, Wen-Tsuen Chen, and Jia-Shung Wang, "Image Sequence Coding Using Adaptive Tree-Structured Vector Quantization with Multipath Searching," in Proceedings of the International Conference on Acoustics, Speech, and Signal Processing, Toronto, Canada, May 1991, pp. 2281--2284.

....node, we pick up many of them, according to some heuristic function, and follow them simultaneously in the lower levels. The choice is iterated at each node down to the leaf level, in which we simply choose the nearest codevector among the selected ones. This technique has been rst introduced in [2] in the framework of video coding with good coding performances improve1 ment. A formal description of the general encoding algorithm is given in algorithm 1. Algorithm 1 A general encoding algorithm for MPTSVQ x is the input vector S is a set of codevector indices T is the tree S f root node ....

....with a xed encoding complexity. 2 3 A PROCEDURE FOR BALANCED MPTSVQ DESIGN The tree in a MP TSVQ should be designed according to a de ned path selection method. The training vectors have to follow exactly the same paths as the encoded vectors, which is not the case in the method described in [2]. The generalization of the classical TSVQ design method to MP TSVQ is quite straightforward. Unlike the standard TSVQ design algorithm, a MP TSVQ can only be designed in a breadth rst way. A set of available paths is associated to each training vector, and updated at each level in the ....

Chen W.T. Chang R.F. and Wang J.S. Image sequence coding using adaptive tree-structured vector quantization with multipath searching. In Proc. IEEE Int. Conf. Acoust., Speech, Signal Processing, pages 2281-2284, 1991.

....the support region based on the previous quantized sample, while in [25] more than one sample of memory is used. Finally in [12] both the support region and the bin sizes can be adjusted, although the bin sizes are restricted to a finite set of values. A somewhat different problem is tackled in [15] where an initial tree structured vector quantizer (TSVQ) is first designed, with a rate higher than the rate available for transmission. Then the adaptive algorithm chooses which subtree of the previously designed tree has to be used at every instant. Both encoder and decoder keep counts of the ....

....assumes that the 165 sources are relatively smooth but have varying dynamic range so that the role of the adaptation is to estimate the changes in the variance of the source (so that the dynamic range of the quantizer is adapted) while a uniform quantizer is used. Similarly, the assumption in [15] is that the initially designed tree structured codebook is sufficiently representative of the expected input signals, so that the adaptive algorithm can find a good sub tree at any given time. In our work, the aim is to explicitly determine a model for the source from the data known to encoder ....

R.F. Chang, W.T. Chen, and J.S. Wang. Image sequence coding using adaptive tree-structured vector quantisation with multipath searching. IEE Proc. I, 139(1):9--14, Feb. 1992.

....Finite state VQ (Chapter 14 of [1] and predictive VQ (Chapter 13 of [1] can also be considered to be local context algorithms. Other reported algorithms of this type include [7] in which conditional probabilities, estimated from surrounding vectors, dictate the choice of local codebook, and [8], in which usage statistics are calculated using a tree organization of the universal codebook. Paul [9] developed a threshold replenishment AVQ algorithm that was one of the first AVQ algorithms ever reported. Paul s algorithm used threshold replenishment with the new codeword being simply the ....

R.-F. Chang, W.-T. Chen, and J.-S. Wang, "Image Sequence Coding Using Adaptive TreeStructured Vector Quantization with Multipath Searching," in Proceedings of the International Conference on Acoustics, Speech, and Signal Processing, (Toronto, Canada), pp. 2281--2284, May 1991.

....reduction in search complexity for large codebook size M . Hence, a one pass adaptive tree structured VQ algorithm would be an interesting and worthwhile entity to pursue. The structure of the codebook, however, poses new complications to the problem. Some previous works of this flavor are [2, 3], which reorganize the structure of the codebook subtree selected from a large pre designed fixed complete tree structured codebook at an update interval. These works do not involve sending of side information. Another direction of future work is a vector quantization scheme with variable vector ....

R. F. Chang, W. T. Chen, and J. S. Wang, "Image sequence coding using adaptive tree-structured vector quantisation with multipath searching," IEE Proc. I, vol. 139, pp. 9--14, February 1992.

No context found.

R. F. Chang, W. T. Chen, and J. S. Wang, "Image sequence coding using adaptive tree-structured vector quantisation with multipath searching," IEE Proc. I, vol. 139, pp. 9--14, February 1992.

No context found.

R. Chang, W. Chen, and J. Wang, "Image sequence coding using adaptive tree-structured vector quantization with multipath searching," IEE Proc. I, vol. 139, pp. 9--14, Feb. 1992.

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