Universal adaptive vector quantization using codebook quantization with application to image compression

A high-resolution analysis is presented for a universal vector quantization scheme based on periodic codebook transmissions. The scheme assumes a slowly changing nonstationary source such as an image and periodically transmits new updated codebooks as side information to the receiver. The side information is transmitted via a large universal codebook which itself acts as a quantizer for the updated codebooks to be transmitted. This scheme generalizes the more simple technique of adapting a codebook by transmitting its vector components one at a time using a fixed uniform scalar quantizer. These schemes are compared both theoretically and experimentally and side information is determined