Abstract: In order to detect and filter the noise more efficiently, the probability filtering algorithm based on the maximum and minimum gray and the directional texture detects the noise based on the maximum and minimum gray. For the suspicious noise of detected, it does second noise detection based on the smooth transitivity of four directional texture. In the filtering window, the signal pixel of highest frequency replaces the noise. If there is no signal pixel in the window, in order to make the window contain signal pixels, the filter window is enlarged. When the filtering window is enlarged to the maximum size, and there is still no signal pixel in the window, the pixel of highest frequency and processed in the window replaces the noise. The proposed algorithm and the current median filtering algorithms, the filtering performance of which is the best, are applied to image filtering experiment. Compared with the current median filtering algorithms, through the comparative analysis of subjective visual effect and objective data of the image filtering results, the probability filtering algorithm based on the maximum and minimum gray and the directional texture has proved that, its filtering performance is better than other algorithms. It maintains well the edges and details of image while filtering effectively the noise.

Key words: image filtering, salt-and-pepper noise, median filtering algorithm, maximum and minimum gray, directional texture, probability filtering algorithm

摘要： 为了更高效地检测和滤除噪声，基于灰度最值和方向纹理的概率滤波算法根据灰度最值进行噪声检测，对检测出来的可疑噪声，根据四个方向纹理的平滑过渡性进行第二次噪声检测。运用滤波窗口中出现频次最高的信号像素取代噪声。如果窗口中不存在信号像素，增大滤波窗口，以使窗口包含信号像素。当滤波窗口增大到允许的最大尺寸时，窗口中依然没有信号像素，则用窗口中已处理的、出现频次最高的像素取代噪声。将算法与当前滤波性能最好的中值滤波算法用于图像滤波实验。从滤波结果的主观视觉效果和客观数据两方面进行的比较分析证明，相对于当前的中值滤波算法，基于灰度最值和方向纹理的概率滤波算法具有更加良好的滤波性能，在滤除噪声的同时，很好地保持图像的边缘和细节。

关键词: 图像滤波, 椒盐噪声, 中值滤波算法, 灰度最值, 方向纹理, 概率滤波算法