Design of high-order conditional entropy coding for images

The recently developed incremental-tree extension technique is used to design the conditional tree for high-order conditional entropy coding. In order to reduce its complexity, the authors introduce two techniques: code table reduction and nonlinear quantization of conditioning pixels. These two techniques substantially reduce the number of code tables, the number of conditioning pixel address-bits, and the number of states with only a minor penalty in performance. The authors also propose a suboptimal method to determine the pixel sequence of the conditioning states. Using these techniques, the high-order conditional entropy coding of images becomes practical and has shown significant improvement over the conventional runlength/variable-length coding scheme