Video residual coding using SPIHT and dependent optimization

We introduce a video residual coding technique based on wavelet transforms and set partitioning in hierarchical trees (SPIHT). In this scheme, wavelet coefficients that correspond to the same spatial location in the image domain are grouped together to form wavelet blocks. Each wavelet block is then converted to a SPIHT-compatible bit stream by an optimized SPIHT encoder minimizing the Lagrangian cost J=D+λR. We also extend the rate-distortion method to include future dependency and propose an iterative algorithm to solve the dependent coding problem. The result is a coding system which achieves better rate-distortion performance than baseline H.263 with a very fast integer-only decoding procedure