Listless zerotree coding using raster tree search

Abstract

Previously, we have proposed a very simple zerotree coding algorithm called listless zerotree coding (LZC) whose coding memory requirement is significantly lower than that of SPIHT. In common with SPIHT, LZC performs a recursive tree search that is likely to discard some important coefficients when the bit rate is low. Therefore, we propose a new LZC algorithm that uses a raster tree search to minimize important information loss. The zerotree structure is embedded into the recursive tree search but for the raster tree search the zerotree structure will need to be stored in the buffer memory. The new LZC utilizes a flag bitmap to store the matrix-wide zerotree structure that is conventionally stored in coordinate lists by other raster tree search zerotree algorithms. Consequently, the new LZC exhibits not only a dramatically low coding memory requirement but also a very low coding complexity.