基于分频算法的VQ码书生成方法  

An Algorithm to Generate VQ Codebook Based on Frequency Dividing

作　　者：陈静瑾[1] 银磊[1] 马文龙[1] 余宁梅[1] 高勇[1] 

机构地区：[1]西安理工大学自动化与信息工程学院,陕西西安710048

出　　处：《西安理工大学学报》2004年第1期31-35,共5页Journal of Xi'an University of Technology

基　　金：国家自然科学基金资助项目(60276017)。

摘　　要：提出一种简单有效的VQ码书生成的分频算法。该算法将图像分割为高频、低频两部分,在高频段采用阈值比较的方法,在低频段采用分块提取的方法生成VQ码书。实验结果表明,该算法生成的码书通用性较强,在码率为0.625bpp的情况下,重构图像的峰值信噪比为23.20～32.24dB,性能优于其他方法。For generating VQ codedbook, a simple and efficient frequency-dividing algorithm is presented in this paper. First, the image is divided into two parts, high and low frequency, then we get the VQ codebook by means of comparing the threshold in high frequency and extracting representative codewords at different regions in low frequency part. The results show that the codebook generated by this algorithm is more universal. The PSNR (peak signal noise ratio) of decoded image at 0.625 bpp is between 23.20 dB and 32.24 dB, and the quality is beter than others obviously.

关 键 词：矢量量化 分频法 峰值信噪比 阈值 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]