彩色数字图像量化的一种新算法

New Algorithm of Color Digital Image Quantization

分析比较了基于方差最小标准和基于最大类内距离最小标准的彩色图像量化算法．方差最小算法着重于最小化整幅图像量化前后的总方差；最大类内距离最小算法着重于减少类内任意点与其代表点的最大距离．文中提出了一种结合两种标准的新算法，在限制类内点与其类代表点最大距离的同时，尽量降低整幅图像量化前后的总误差．经验证。

Two kinds of Vector Quantization (VQ) algorithms were discussed. One is based on minimizing variance and the other is based on minimizing maximum discrepancy. The two algorithms have different emphasis and results. The former is to minimize the total variance between the original and quantized image but offer no guarantees regarding the maximum color error; while the latter is to minimize the maximum discrepancy but offer no guarantees regarding the total variance. Therefore, a new algorithm was suggested combined the two criteria. Compared with the other algorithms, the results of this new algorithm are more satisfactory and competitive.