Algorithm and VLSI Design of a Feature-Based Classified Vector Quantizer for Image Coding
|
Abstract:
In this paper, a feature-based classified vector quantiza-tion (FCVQ) algorithm and VLSI implementation of the classifier are presented. The FCVQ technique exploits the characteristics of discrete cosine transform (DCT) and the concepts of block truncation coding (BTC) to simplify the classification and preserve the edge information efficiently. During the classification, an input block is classified as an uniform block or a block with a specified orientation of edges at first; different partition schemes are used for encoding different types of blocks. There are only simple additions required for the classification. Simulation results show that the approach alleviates edge degradation in pri-mary VQ with bit-rates as low as 0.35 bits per pixel (bpp). The proposed classifier has been implemented by 0.8-/m CMOS technology on a 6.3 mmx 4.7ram die with a clock rate of 35 MHz.
Citations
| 827 | R.: An algorithm for vector quantization design – Linde, Buzo, et al. - 1980 |
| 52 | Classified vector quantization of images – Ramamurthi, Gersho - 1986 |
| 26 | Image compression with variable block size segmentation – Vaisey, Gersho - 1992 |
| 2 | Systems for Video Tech – Circuits - 1992 |
| 2 | Variable-rate contour-based interpolative vector quantization for image coding – Ho, Gersho - 1988 |
| 2 | Modular VLSI Architectures for real-time full-search-based vector quantization – Park, Prasanna - 1993 |
| 1 | Tree-Structure Architecture and VLSI Implementation for Vector Quantization Algorithms – Jong - 1994 |
