Adaptive Tree-Structure Lattice Vector Quantization for Video Coding
Résumé
The purpose of this paper is to introduce a new adaptive vector quantization (AVQ) for the compression of digital image sequences. In our previous studies we proposed a lattice vector quantizer (LVQ) based on the hierarchical packing of embedded truncated lattices. Now we investigate the capability of this LVQ for AVQ, through a scheme taking account of its tree-structure. Precisely, in order to fit the spatiotemporal statistics of the image sequences, the codebook is designed using a training procedure and in two parts, with: a stump from several types of training sequences; some branches added according to the sequence to be coded. Experimental results are given with the adaptive LVQ taking place in a subband coder.