Abstract: When dealing with the high-contrast images contaminated by salt-and-pepper noise, the traditional Block-Matching and 3D filtering（BM3D） can not effectively preserve the edges and texture details of the image, it will appear edge-ringing effect in the denoised image edges. In order to improve the shortcomings of traditional BM3D algorithm when dealing with salt and pepper noise, an improved denoising algorithm of BM3D is proposed, which uses the edge direction instead of the horizontal direction to search for similar blocks. The experimental results show that the number of similar blocks obtained by the improved BM3D algorithm is three times higher than that of the traditional BM3D algorithm, and the Peak Signal-to-Noise Ratio（PSNR） is further improved, and the edge of the image is effectively preserved while salt and pepper noise is removed.

Key words: salt and pepper noise, Block-Matching and 3D filtering（BM3D）, similar block, image denoising

摘要： 在处理由椒盐噪声污染的高对比度图像时，使用传统的三维块匹配算法（Block-Matching and 3D filtering，BM3D）去噪不能有效保留图像的边缘和纹理细节，在图像的边缘会出现边缘振铃效应。为了改善传统BM3D算法在处理椒盐噪声时的不足，提出了用边缘方向代替水平方向搜索相似块的BM3D改进去噪算法。实验结果表明，改进BM3D算法获得的相似块数量是传统BM3D算法的3倍，峰值信噪比（PSNR）也得到进一步提高，在去除椒盐噪声的同时也使图像边缘得到有效保留。

关键词: 椒盐噪声, 三维块匹配（BM3D）, 相似块, 图像去噪