高效的模糊聚类初始码书生成算法

码书设计在矢量量化中至关重要,而多数码书设计算法都是基于初始码书的.从经典的LBG算法的缺陷出发,提出一种基于模糊聚类的高效初始码书生成算法,通过将初始码书的码矢在输入矢量空间中很好地散开,并尽可能占据输入概率密度较大的区域,从而使之后的LBG算法避免陷入局部最优,设计出的码书性能更好,更加接近全局最优,同时加快了收敛速度,减少了迭代次数.将该算法应用于图像编码的实验中,结果表明:该算法能够从效率和质量两方面有效地提高矢量量化的性能.

Novel and efficient generation algorithm of initial codebook based on the fuzzy clustering theory

Design of the codebook is of great importance in the vector quantization,and most algorithms of designing codebook are based on the initial codebook.From the drawbacks of the classical LBG algorithm,a novel and efficient generation algorithm of the initial codebook based on the fuzzy clustering theory was presented.With this algorithm,the code vectors of the initial codebook could be dispatched well in the vector space,and the area,whose input probability density was larger,was occupied.After that,the LBG algorithm could avoid being trapped in the local optimization and the codebook would be in better performance and closer to the global optimization with faster convergent speed as well as fewer times of iteration.This novel algorithm was applied to the image codin gexperiment,and the result shows that it can enhance the performance of the vector quantization,both in efficiency and quality.