Optimal Shape Coding under Buffer Constraints

In this paper, we consider an optimal rate control algorithm for shape coding of emerging MPEG-4 compression system. Current MPEG-4 shape coding consists of two sub-steps; First distortion is introduced by scaling down and up way. Then, second, context-based arithmetic coding is applied. Since the arithmetic coding is \lossless", scaling down and up step is considered as a virtual quantizer. We rst formulate the bu erconstrained adaptive quantization problem for the shape coding. And then we propose an algorithm for the optimal solution under the bu er constraints. Experimental results are given using MPEG-4 shape coder. 1. OPTIMAL BUFFERED COMPRESSION Recently \optimal Trellis-based bu ered compression" was proposed in [2] to provide optimal bu er control strategies for signal sequences in a nite bu ered environment. Authors, rst, formalized the description of the bu er-constrained adaptive quantization problem. And, second, they formulated the optimal solution for a given set of admissible quantizers used to code a discrete nonstationary signal sequence in a nite bu er. Optimal bu ered compression is de ned as followings. Problem De nition: Given a set of quantizers, a sequence of blocks to be quantized, and a nite bu er, select the optimal quantizer for each block so that the total cost measure is minimized and the nite bu er is never in over ow. Consider the allocation for N blocks, and suppose there are M quantizers available to code each block. Let dij and rij be, respectively, the distortion and the number of bits produced by the coding of block i with quantizer j, and let r be the channel rate per block. Dene an admissible solution x as a selection of one quantizer for each block, i.e., a mapping from 1; 2; :::; N to 1; 2; :::;M, x = fx(1); x(2); :::; x(N )g, where each x(i) is e