Improved and optimal adaptive hierarchical vector quantization for image coding and indexing

Adaptive hierarchical algorithms of vector quantization (VQ) for image coding and database applications are proposed. The basic codebook is generated adaptively using adaptive VQ; then the quadruplets of codes/indices are coded into higher-level codes by creating higher-level index-codebooks to reduce the redundancy presented in the codes. A partial encoding scheme is proposed in the second layer coding to improve coding efficiency. Optimal design of the higher-level codebook is discussed and proved to be NP-complete. A pseudo-optimal algorithm is proposed instead. These schemes achieve a much better compression ratio with no extra distortion and little extra computation than the single layer adaptive VQ. Applications of these schemes to content-based information retrieval in multimedia database systems are discussed.