We introduce a new wavelet image coder which is the forward-adaptive counterpart of the state-of-the-art backward-adaptive Estimation-Quantization coder. The variance field associated with the wavelet coefficients is estimated, lossily compressed, and transmitted as side information. Next, the wavelet coefficients are compressed using that side information. We optimize the resulting two-part coder based on an informationtheoretic analysis. The performance of the forwardand backward-adaptive compression algorithms is compared based on several numerical experiments.
Anthony F. Docimo, Kannan Ramchandran, Mehmet Kiva