Variable region vector quantization

Abstract

AbstractGeneralized algorithms for vector quantization are presented and their convergences are proved. The generalized vector quantization is called variable region vector quantization since the method allows adjusted variable dimensional vectors covering variable subregions of source data.Algorithm I is the generalization of the LBG algorithm into the variable region case. This is called full‐gain variable region vector quantization. Algorithm II, on the other hand, is the variable region generalization of the gain‐shape type. The formation of each variable subregion is due to the grouping of elements so that the resulting set of variable dimensional super‐vectors has the minimum distortion to a given codebook. Algorithm III considers encoding‐decoding. Algorithm IV gives the suboptimal minimization for the alleviation of computational load.Methods presented here are applicable to various pattern handling including artificial neural nets. However, data compression terminologies are adopted in this research. All of the foregoing algorithms are introduced without any physical entity to data. Later, the cases of speech and images are discussed using their specific natures. The examples of the region optimization on speech and images are given.