Hierarchical image vector quantizer using gradient-Laplacian subspace distortion

A new hierarchical mean-residual image vector quantization (HMRIVQ) technique using a general four-dimensional gradient-Laplacian subspace distortion (GLSD) is proposed. The GLSD is used for design and implementation of different block sizes of the HMRIVQ. Therefore, the computational and storage costs of the encoder are significantly reduced. The HMRIVQ can also achieve a very high compression rate for low detailed images. A performance comparison of the HMRIVQ using Euclidean distortion and the general four-dimensional GLSD is also given