采用分量均值正交分割法设计矢量量化初始码书-中南大学学报.PDF

41 2 () Vol.41 No.2 2010 4 Journal of Central South University (Science and Technology) Apr. 2010 ( 410083) 1.0~1.3 dB30% 0.1 dB1.0~1.5 2 TN911.73 A 1672−7207(2010)02−0628−04 Initial codebook of vector quantization designed by component mean orthogonal segmentation DENG Hong-gui, GUO Sheng-wei (School of Physics Science and Technology, Central South University, Changsha 410083, China) Abstract: A new algorithm component mean orthogonal segmentation algorithm (CMOSA) was proposed to generate vector quantization initial codebook using component mean. The results show that in different images and different codebook sizes, compared with random algorithm, peak signal to noise ratio rises by 1.0−1.3 dB, and training time increases no more than 30%. Compared with splitting algorithm, peak signal to noise ratio difference is less than 0.1 dB, but the training speed improves 1.0−1.5 times. Between training speed and quantization distortion of the two conflicting indicators, CMOSA achieves a better balance than other algorithms. Key words: vector quantization; initial codebook; component mean; orthogonal segmentation (Vector quantization, VQ)[1] N M 1 Linde [2]LBG M LBG M