摘要： 提出一种保边缘的去噪方法。该方法利用4个方向核卷积检测图像中的噪声点和非噪声点，对非噪声点不进行处理。在噪声点的3×3邻域中选择距离最小的非噪声点，若无非噪声点则邻域扩大为5×5，若有非噪声点则用这些点的中值替换噪声点，否则用邻域中距离最远像素点的均值替换噪声点。实验结果表明，该方法不仅能有效去除图像的噪声，且能较好地保护图像边缘。

关键词: 卷积, 噪声点, 非噪声点, 邻域, 保边缘

Abstract: This paper proposes a edge-preserving image denoising method. The noise point and noise-free point are computed by four directional convolutions operators, and noise-free point are not dealed with. The mininum distance noise-free point in the 3×3 neighborhood is obtained. If there is not noise-free point, the neighborhood is extended to 5×5. The median value of them is used to replace the noise point while existing noise-free point. Otherwise, the noise point is replaced by the average value of the maximum distance pixel. Experimental result shows that the algorithm can effectively remove image noise, and can better protect the image edge.

Key words: convolution, noise point, noise-free point, neighbourhood, edge-preserving

中图分类号: 

• TP391