Fast and accurate rate prediction and picture quality control for wavelet image coding

By introducing the concepts of characteristic rate curves and rate curve decomposition, a novel framework for rate-distortion (R-D) analysis and source modeling is developed in this work. With this framework, a fast algorithm is proposed to accurately estimate the R-D curve before quantization and coding for wavelet-based image coders. The proposed algorithm is applied to the SPIHT (set partitioning in hierarchical trees) and stack-run (SR) encoders. Our extensive experimental results show that the relative R-D curve estimation error is less than 5%.