A low complexity multiresolution approach to image compression using pruned nested tree-structured vector quantization
S.M. Perlmutter, Robert M. Gray · 2002
A novel algorithm is described for constructing a progressive, multiresolution compression code. The codec consists of nested levels of tree-structured vector quantizers (TSVQs) where the codebook for each level of the nested TSVQs is constructed from the terminal leaves of the TSVQ from the previous level. In order to generate a multiresolution output in a progressive manner, the codeword dimension at each level's TSVQ is greater than or equal to those of the previous levels. Pruning is performed on the nested TSVQs to achieve the bit allocation across the levels. The resulting pruned TSVQ provides a multiresolution output with low computational complexity at the decoder while simultaneously providing superior performance to ordinary pruned TSVQ at low bit rates.>