Adaptive vector quantisation of non-orthogonal representations for image compression

Abstract

An efficient compression technique employing adaptive vector quantisation of multiple non-orthogonal transform domain representations of still images is developed. For each sub-image, the encoder selects a code from the domain that yields best representation. The performance improvement employing the proposed technique relative to existing single domain vector quantisation coding methods, for the same compression ratio, is obtained at the expense of increased computational complexity.