Abstract: Against salt-and-pepper noise in gray image, this paper proposes a method to distinguish between noise and non-noise by absolute grey relation degree, and uses median filtering algorithm to denoise the noise. After combining the pixels in [n×n]（[n] is an odd number greater than or equal to 3） template into two groups, the paper uses absolute grey relation degree analysis algorithm to calculate the similarity of these two pixels groups, then it can know whether the current pixel is noise or not. For the noise pixel, denoise it by median filter；for the non-noise one, do nothing, thus retain good image detail. Experimental results show that, this method has good denoising results compared with similar methods.

Key words: analysis of grey absolute relation, median filter, relation coefficient, salt-and-pepper noise

摘要： 针对灰色图像中的椒盐噪声，提出了一种利用绝对灰色关联度分析方法区分噪声与非噪声，并结合中值滤波进行去噪的算法。将[n×n]（[n]为大于或等于3的奇数）模板中的像素组合成两组序列，利用灰色绝对关联分析法，计算出这两组序列的相似关联度，从而判断当前像素是否为噪声。对于被判定为噪声的像素，进行中值滤波处理，以实现去噪；对于非噪声像素则不作处理，从而较好地保留了图像的细节。实验结果表明，与同类方法相比较，此方法有良好的去噪效果。

关键词: 绝对灰色关联分析, 中值滤波, 关联系数, 椒盐噪声