Lossless multi channel predictive coding for images

This report proposes a new lossless coding of images. The decoder can expand a rough image from a part of the bit stream (compressed data) and also it can expand the original image from the rest. This functionality is useful for progressive transmission of images or browsing images in a huge database. The new method is based on a reversible wavelet (RWT) and a lossless multi-channel prediction (LMP). Filter coefficients of the LMP are optimized for each input image so that the total bit rate is minimized. Experimental results indicate its effectiveness for use of lossless image coding. Entropy rate is reduced by 0.35 bpp in average for some images.