Abstract: This paper proposes a two-stage color quantization method. At the first stage, a palette selection scheme suitable to the quantization level requirement is chosen and an initial palette is selected. At the second stage, LBG algorithm is adopted to iteratively refine the palette. Experimental results show that the scheme is effective for reducing both iterations needed and the final MSE. The results also prove the approach is superior to most of the prevalent methods in terms of quantization distortion measured by the MSE metric.

Key words: Color quantization, Mean squared error, Popularity scheme, Merging scheme, LBG algorithm

摘要： 提出了一种基于LBG的二阶色彩量化方法：第1阶段根据量化级别的大小确定初始调色板选择方案并遴选出初始调色板；第2阶段利用LBG算法对调色板进行迭代优化。实验结果表明，该方案能明显地减少LBG收敛所需的迭代次数以及最终的量化失真。相对于当前其它量化方法，该方法具有更小的量化失真。

关键词: 色彩量化, 均方误差, Popularity准则, 合并准则, LBG算法

CLC Number: 

• TP18