Image Coding with Finite State Vector Quantization Utilizing Extrapolation Selection
A novel extrapolation selection finite state vector quantizer (ESFSVQ) has been devised. The state codebook selection algorithm is based on linear extrapolation predictions from the immediate upper and left neighboring blocks of each incoming source vector. It may be characterized as a vector quantizer with memory. The main accomplishment of ESFSVQ is to assure access to a large master codebook for quantizing purposes, and thus achieve high image quality, while utilizing the low bit rate of a small state codebook, for the purpose of specifying the block label. A master codebook of 4 × 4 blocks of size 256, 512 or 1024, state codebook sizes of 32, 64, or 128 vectors, was employed in implementing ESFSVQ. The results have shown that high image quality can be achieved, upon image reconstruction, at coding bit rates of 0.30-0.45 b/pixel.
MSU Digital Commons Citation
Manikopoulos, Constantine N.; Sun, Huifang; and Antoniou, George, "Image Coding with Finite State Vector Quantization Utilizing Extrapolation Selection" (1989). Department of Computer Science Faculty Scholarship and Creative Works. 329.