LIFTING-BASED DESIGN OF REVERSIBLE CELLULAR AUTOMATA FOR SCALABLE CODING OF BINARY IMAGES

Reversible cellular automata have been proposed as non-linear filter banks for spatially scalable binary image compression. Previous design approaches led to efficient coders for simple images but do not guarantee detail preservation across scales. In this work we are concerned with the design of optimized cellular automata rules that offer better performances. The lifting principle is used for guaranteeing reversibility. We obtain a scalable coder that (i) offers the same quality of JBIG in the reduced scale images, (ii) achieves high compression ratios on simple images, (iii) provides at least the same performance of JBIG on any image content, and (iv) employs arithmetic coding with a very small number of contexts for reduced complexity.