Prediction-Based Compression Ratio Boundaries for Medical Images

Most present prediction-based image compression techniques take advantage of either intraor inter-image correlation or both to de-correlate images. These decorrelated images usually contain reduced information content. Consequently, higher compression ratios may be obtained from these de-correlated images. This paper studies the relationship between image correlation and the resultant information redundancy, and develops a method for estimating the compression ratio C , which is a function of the image correlation ρ . The compression boundaries (i.e., the maximum compression ratios) are further derived from this compression ratio function. The prediction-based compression technique has been applied on some magnetic resonance (MR) brain image sets to numerically prove the derived compression boundaries.