Title :
A vector quantizer with short-time memory for digital image coding
Author :
Saito, Takahiro ; Komatsu, Takashi ; Harashima, Hiroshi
Author_Institution :
Dept. of Electr. Eng., Kanagawa Univ., Yokohama, Japan
Abstract :
The authors propose a novel scheme of vector quantization with memory, named vector quantization with short-time memory (VQ-STM). The VQ-STM has two types of codebooks, a fixed-structure codebook whose codewords are kept unchanged, and a variable-structure codebook, which is defined as a subset of the fixed-structure codebook and whose structure is continually updated depending on the past behavior of the coder. The VQ-STM coder chooses the proper codebook between two codebooks for every input vector. The update of the variable-structure codebook is easily performed only by simple heuristics for maintaining a finite-length self-organizing sequential list which stores past coder outputs. The results of experimental simulations conducted on monochrome images demonstrate that significant compression gains can be achieved by enhancing the vector-quantization approach with the concept of short-time memory.<>
Keywords :
encoding; picture processing; VQ-STM; codebooks; compression gains; digital image coding; finite-length self-organizing sequential list; fixed-structure; heuristics; input vector; monochrome images; short-time memory; variable-structure; vector quantizer; Algorithm design and analysis; Clustering algorithms; Compaction; Computational complexity; Decoding; Digital images; Image coding; Partitioning algorithms; Rate-distortion; Vector quantization;
Conference_Titel :
Circuits and Systems, 1988., IEEE International Symposium on
Conference_Location :
Espoo, Finland
DOI :
10.1109/ISCAS.1988.14906