Preferential multi-objective genetic algorithm for JPEG quantization table optimization

This paper proposes a novel multiobjective genetic algorithm to optimize both the luminance and chrominance quantization tables for JPEG compression. One of the key factors influencing the performance of JPEG compression is the quantization table. The compression ratio and the decoded image quality are determined simultaneously by the quantization table. Optimizing both the compression ratio and decoded image quality through the quantization table is a multi-objective problem by its nature, and there is always a trade-off between these two performances. The multi-objective genetic algorithm is very suitable to solve this kind of problem. Therefore, a preferential NSGA-II selection mechanism is proposed to optimize the quantization tables and hence improve the JPEG compression performance. The experimental results indicate that the proposed approach can generate diversified optimal quantization tables for JPEG compression, some of which are even better than the traditional quantization tables specified in the JPEG standard for both the decoded image quality and compression ratio.