Adaptive Down-Scaling Techniques for JPEG-Based Low Bit-Rate Image Coding

The DCT-based JPEG standard remains to be the most popular compression utility for digital images despite the emerging wavelet-based JPEG-2000 standard. A commonly recognized weak point of the DCT-based transform coding is its blocking effects which become increasingly visible when the bit-rate requirement goes deeper into the low bit-rate territory. There has been a great deal of research endeavors to deal with this problem. Motivated by a recent work of Bruchstein, Elad, and Kimmel (BEK), in this paper we propose two adaptive techniques that can be applied to a BEK type of transform coding system for further performance improvement at a cost of moderate increase in complexity. Here the adaptation is based on the input image's local features and statistics, and goes in two ways: either adaptive in terms of the rate of the down-sampling or in terms of the quality factor of each image block. Simulation results are presented to demonstrate the effectiveness of the proposed techniques