【摘要】 向量量化是图像压缩中的重要内容,而码书生成是向量量化的关键.提出了一个全新的、简单的码书生成算法,其基本思想是以向量量化聚类性质为基础,应用图论建立数据之间的离散关联设计算法.该算法与传统的算法相比,优势在于不需要初始码书,不需要在实际应用中几乎不可能知道的高维向量集合的概率分布,不需要Voronoi划分,同时它避免了一般算法局部最优问题.更多还原

【Abstract】 Vector Quantization is very important in the image compression. The generation of the codebook is critical in Vector Quantization. A novel and simple codebook generation algorithm for VQ is presented in this paper. The basic idea of the algorithm is that the discrete relationship of the data is constructed to design the corresponding algorithm by Graph Theory according to the clustering property of the Vector Quantization. Comparing with the traditional algorithms the new algorithm has the following difference: the traditional algorithms are designed by vectors, distance, and transform, such as FT transform and Wavelet transform. The new algorithm does not need initial codebook, the distribution of vectors and the Voronoi partition. It can evade local minimum problem.更多还原