基于局部直方图的加权均值滤波器

Weighted mean filter based on local histogram

针对受椒盐噪声不同程度污染的灰度图像滤波问题,提出了一种基于局部直方图的加权均值滤波算法。该算法针对椒盐噪声特点,对含噪图像进行噪声检测,建立噪声标记矩阵;对标记为信号的像素不做处理,标记为噪声的像素依据其邻域像素污染程度采取不同窗口尺寸的加权均值滤波,像素权值由噪声点所在区域的局部直方图确定。对不同噪声率污染的测试图像仿真结果表明,该算法在有效抑制噪声的同时能较好地保持图像的细节信息。最后,通过与中值滤波和其他一些改进算法的滤波效果比较,证明了该算法的有效性。

A weighted mean filtering algorithm was proposed for the filtering problem about gray-scale image that was polluted by sale and pepper noise of different degree. According to the characteristic of salt and pepper noise, the algorithm detected image noise, established noise marked matrix, without processing the pixels marked as signal. As for the pixels marked as noise, weighted mean filtering based on degree of pollution of its neighbor pixels with different window size was carried out. The weight of pixel was determined by the local histogram of noise point region. The result of image simulation shows that the algorithm can suppress noise efficiently and preserve the image detail information. Finally, the comparison of weighted mean filtering and other improved algorithms proves the effectiveness of the algorithm.