Abstract: In order to improve the filtering effect of noise image, this paper put forward a new filtering algorithm. This algorithm consisted of three stages. Firstly, the similarities of the pixels were used in the image to detect the impulse noise. Then the filter window was divided into eight main directions to determine the directions of the edges, and at last these impulse noises were restored using an edge-preserving method. The simulation results indicate that this algorithm can not only accurately detect the noise points, but aslo protect the noise-free pixels and the boundaries in the noise image when the noise density is small.

Key words: similarity noise detection, edge-preserving filter, impulse noise, image denoising

摘要： 为了改善受脉冲噪声污染的图像的滤波效果,提出了一种新的滤波算法。该算法包括3个阶段,首先，利用像素点之间的相似性来检测图像中受噪声污染的像素点;然后，将滤波窗口分为8个主要方向来确定边缘方向;最后，针对噪声点进行边缘保护滤波。实验结果表明,在噪声污染度较小的情况下,该算法不仅能准确地检测出噪声点,而且更多地保护了噪声图像的边缘部分以及非噪声点,具有良好的滤波效果。

关键词: 相似性噪声检测, 边缘保护滤波, 脉冲噪声, 图像降噪