An efficient pyramid image coding system

The pyramid image structure can be naturally adapted for progressive image transmission over low-speed channels and hierarchical image retrieving in computerized image archiving. An efficient pyramid image coding system using quadrature mirror filters to form the image-pyramids is proposed in this paper. Characteristics of the image-pyramids are presented. Since the Laplacian pyramids of most nature images contain sparse and spatially concentrated data, a combined run-length coding for zero-valued elements and entropy coding for elements larger than a certain threshold is employed. The textural features in the Laplacian pyramids suggest that coding techniques pursuing spatial correlation may be advantageous. Therefore, vector quantization is chosen to code the Laplacian pyramids. Simulation results have shown that simple vector quantization accomplished significant bit-rate reduction over scalar quantization. The proposed system has also shown good-quality reproduction at bit rates lower than 1 bit/pixel.