一种可调节的多级颜色量化算法

针对当量化图像的颜色数量较少时如何调节颜色层次和微弱颜色,本文提出了一种新颖的多级颜色量化算法.根据Kohonen自组织神经网络和线性像素置换计算第一级调色板;然后多次迭代找出最相似的颜色对,删除其中的一个,得到第二级调色板;最后根据调色板中颜色的像素数量和与基色距离的加权和,选出最终的代表颜色.实验结果表明,该算法能够通过设置适当的像素取样率、像素数量权重和与基色的距离权重而有效地选择满足不同需要的量化结果,并且保持较好的量化图像质量.

An Adjustable Multi-level Color Quantization Algorithm

This paper proposes a novel multi-level algorithm for color quantization,on how to make a choice between weak colors and color layers when a color image is reproduced with very small number of colors.Firstly,an initial color palette is constructed based on Kohonen Self-organizing Neural Network and Linear Pixel Shuffling;then repeatedly the two most similar colors are found in the initial palette,and one is eliminated,until only the required number of colors are left;finally,the weighted sums of pixel numbe...