EXPERIMENT PERFORMANCE ANALYSIS ON SENSITIVITY TO INITIAL CODEBOOK OF LBG
Wang Mao-zhi
Abstract
Wang Mao-zhi
Abstract
The basic concepts of vector quantization and its LBG~([1]) algorithm are introduced, and the defects of LBG are analysed in this paper. With three different initial codebook design methods, this paper verified the conclusion that the LBG algorithm is sensitive to the initial codebook. At the same time, the experiment performance and a method on describing the initial codebook as well as an equal interval method on designing initial codebook is also proposed. At last, some improved methods aimed at the defects of the sensitivity to the initial codebook of LBG algorithm are presented in the paper.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
The basic concepts of vector quantization and its LBG~([1]) algorithm are introduced, and the defects of LBG are analysed in this paper. With three different initial codebook design methods, this paper verified the conclusion that the LBG algorithm is sensitive to the initial codebook. At the same time, the experiment performance and a method on describing the initial codebook as well as an equal interval method on designing initial codebook is also proposed. At last, some improved methods aimed at the defects of the sensitivity to the initial codebook of LBG algorithm are presented in the paper.
Key concepts: Codebook, Linde–Buzo–Gray algorithm, Sensitivity (control systems), Vector quantization, Algorithm, Computer science, Mathematics, Mathematical optimization