An Improved VQ Codebook Search Algorithm Using Principal Component Analysis
✍ Scribed by Chin-Chen Chang; Dai-Chuan Lin; Tung-Shou Chen
- Publisher
- Elsevier Science
- Year
- 1997
- Tongue
- English
- Weight
- 532 KB
- Volume
- 8
- Category
- Article
- ISSN
- 1047-3203
No coin nor oath required. For personal study only.
✦ Synopsis
process of VQ is achieved by using only the indices of the closest codewords for storage and transmission.
We present an improved codebook search algorithm in this paper. We call it the double test of principal components
As is obvious, choosing the closest codeword for each (DTPC). This algorithm speeds up the codebook search by input vector is the most time-consuming factor in the ensearching only an appropriate sub-codebook instead of the coding process. In order to search the closest codeword in whole set of codewords. Moreover, DTPC inherits several benethe codebook for each vector of an image, an exhaustive fits from some previous techniques, such as the double test search is required. We call this method the full-search (FS) (DT) and the principal component analysis (PCA). Thus DTPC method. The FS obviously requires a large amount of comis much more efficient than the other algorithms. Simulation putation to evaluate the distortion between the input vecresults confirm this efficiency. According to these results, the tor and its nearest codeword. That is, FS needs a considertotal number of the mathematical operations needed in DTPC able amount of time to do encoding. Thus several is usually less than that needed in any other method, even algorithms [1-3, 5-14] have been designed to reduce the if the MSE degradation of DTPC is limited within 0.13 dB. Furthermore, in some cases, this number of DTPC is only 3% computational work and to accelerate the encoding proof that in a full search. © 1997 Academic Press cess. Among them, tree-structured vector quantization (TSVQ) [10,11], double test (DT) method [9], eigenvector method (EVM) [2], partial distortion elimination (PDE)