Super predictive-transform residual vector quantization

A superimposed (Super for short) minimum mean squared error (MMSE) predictive-transform (PT) residual vector quantization (RVQ) coder is introduced in this paper. The Super MMSE PT-RVQ coder arises as a straight forward generalization of a recently offered MMSE PT-RVQ coder and addresses in a simple and elegant fashion the noticeable blocking artifacts that are encountered with the MMSE PT-RVQ coder when operating at very low bit rates. The technique is illustrated by compressing five monochrome test images obtained from the JPEG suitcase at compression ratios as high as 229 to 1 or equivalently at bit rates as low as 0.035 bps. It is found that for all test images the Super PT-RVQ coder generally outperforms the perceptual quality and peak SNR derived with the Said and Pearlman wavelet algorithm.