Adaptive interleaved vector quantization in transform domain

A novel image coding scheme, adaptive interleaved vector quantization (AIVQ) in the transform domain, is proposed. The basic idea of this algorithm is that the original image is first partitioned into large blocks. Each block is then further decomposed into four interleaved, nonoverlapping subblocks, which are subsequently encoded using vector quantization in the transform domain. All or some of the four subblocks are selected to be directly coded by vector quantization according to the local activity of the large block. The subblocks that are not directly assigned to be vector quantized will be calculated from the coded data. Since a strong correlation exists among the subblocks of the blocks with low local content activity, the approximation will not result in large error. Therefore, the bits saved from the lower activity area are allocated to the higher activity area. A variable-rate coder that delivers a constant image quality is obtained