A pyramid image coder using classified transform vector quantization

This paper describes a progressive image coding scheme that employs a discrete cosine transform-classified vector quantization (DCT-CVQ) in a hierarchical pyramid structure. This paper also proposes a new method for designing filters to construct an image pyramid. In designing the filter, by minimizing the maximum error energy between the input and the actual decimated-interpolated signals the optimal half band FIR filter for pyramid construction is described. Computer simulation results show that the proposed half-band filter yields a higher peak signal-to-noise ratio (PSNR) as well as better image quality, in comparison with other currently used filters of the same computational complexity. Compression is achieved by applying the DCT-CVQ coding scheme to each stage. Since a classification is based upon the AC energy of the subblock, progressive transmission can be achieved by transmitting from the topmost to the lowest level of planes in the Laplacian pyramid, and from the higher to the lower AC energy class in the plane. The computer simulation demonstrates that the reconstructed image by the proposed scheme shows a high PSNR (33.78 dB) as well as good visual quality at 0.409 bit per pixel (bpp).