SIMULATED ANNEALING APPLIED TO OPTIMAL DCT QUANTIZATION

We present a method for determining optimum quantization tables for us e in image compression systems which conform to the ISO/CCM standard f or Image Compression, also known as the Joint Photographic Experts Gro up (JPEG) standard1. An algorithm based on simulated annealing2 compre sses and decompresses any collection of 8 by 8 pixel blocks, while sea rching the space of 8 by 8 quantization tables for optimum fidelity ac cording to some chosen measure. A composite cost function maintains a predefined compression ratio while minimizing the RMS error in the dec oded image compared to the original. The process can be carried out on a Taw image, or greater weight can be given to a selected range of DC T coefficients on the basis of psychophysical considerations3. The res ults of applying the methods to the intensity (y) components of three JPEG test images are presented. In all cases improved fidelity as meas ured by RMS error is obtained compared to the quantization table sugge sted in the JPEG standard. Significantly, the quantization tables obta ined for one image most often provide smaller error when applied to ot her images than does the table suggested with the JPEG standard. An un expected result when using pre-emphasized images suggests that the psy chophysical assumptions underlying the suggested JPEG table may be ove rsimplified. By applying the method to blocks selected from a variety of images, improved quantization tables can be found for images in gen eral, including color.