Abstract: Fast peer group filter utilizes the cardinality of peer group in filtering window to judge whether the central pixel is noisy or not. To improve its computation efficiency, the paper introduces a new fast method based on peer group subset in HSL color space. The new approach exploits the hue, saturation and lightness color component and only takes the subset of original filter windows into the judgement of noise existence. The experimental results demonstrate that the new filter can reduce half of computation time at least meanwhile keeping the original processed image quality.

Key words: color image, vector median filter, impulsive noise, switching filter

摘要： 快速对等组矢量滤波通过计算滤波窗口内的对等组的势来判断中心像素是否为噪音。为进一步提高其计算效率，提出一种基于HSL颜色空间的对等组子集的快速滤波方法。该方法利用HSL颜色空间的色度、饱和度和亮度等分量，使用原滤波窗口子集获得的对等组子集来判断噪音的存在。实验结果证明该开关型滤波方法较原对等组滤波方法在保存图像处理效果的同时减少至少一半以上的计算量。

关键词: 彩色图像, 矢量滤波, 脉冲噪音, 开关型滤波