Image compression using learning vector quantization

The authors present a study and implementation of still image compression using learned vector quantization (LVQ). Kohonen's self-organizing feature map (SOFM) has been used to compress several still monochrome images by a compression ratio of 16:1 at a compression rate of 0.5, while maintaining a peak signal-to-noise ratio (PSNR) of about 30 dB. C programs were written to implement learning, compressing, decompressing, analyzing error, and others for the VQ method. These programs were run on the SUN SPARC Station 2. Methods for optimizing learning are presented. Given an image that is subjectively similar to the training image, and if the histogram of the test image is a subset of the histogram of the training image, then quantization of the test image will produce results comparable with the PSNR achieved by the training image. >