Performance analysis of a vector quantizer with cache memory

Applies the cache memory scheme for an efficient implementation of a vector quantization system. The authors structure their vector quantizer to have a large "main" codebook and a smaller "cache" codebook. The main codebook holds all of the possible code vectors for the quantizer...

Full description

Saved in:
Bibliographic Details
Main Authors: Choo, C.Y., Cicchetto, B.A., Nasrabadi, N.M.
Format: Conference Proceeding
Language:English
Subjects:
Cache memory
Image coding
Image quality
Performance analysis
PSNR
Radio access networks
Vector quantization
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Applies the cache memory scheme for an efficient implementation of a vector quantization system. The authors structure their vector quantizer to have a large "main" codebook and a smaller "cache" codebook. The main codebook holds all of the possible code vectors for the quantizer, while the cache codebook holds only those code vectors that have been used recently and are expected to be used again soon. The authors ran simulations for a fully associative cache codebook with least-recently-used replacement algorithm. Cache codebook hit ratios ranging from 60% to 90% were achieved with PSNR of 33-36 dB for line sizes of 1 to 8 and code vector dimension of 4. This scheme improves the encoding performance without significantly affecting the compression ratio and image quality.< >
ISSN:1058-6393
2576-2303
DOI:10.1109/ACSSC.1993.342428