Abstract: An efficient kind of filtering algorithm is given to restore images corrupted by salt-pepper impulse noise.First it selects a 3×3 neighborhood.According to whether gray value of arbitrary pixel is its neighborhood’s extremum,all pixels are divided into doubtful noises pixels and signal pixels.Then it constructs 8 templets of orientation gradient for each doubtful noise,makes the sure of the noises sets according to the magnitude of 8 values of orientation gradient.The results of experiment prove that the new method not only saves the original image information and has a good Signal-to-Noise Ratio,but also removes noises simultaneously.

Key words: salt-pepper noise, orientation gradient, extreme value, median filter

摘要： 提出了一种有效的椒盐噪声图像滤波算法。该算法首先对含有噪声的图像取3×3邻域,判断某点是否为邻域极值将全部像素点分为可疑噪声与信号点集合;其次对每一个可疑噪声点在其3×3邻域内构造八个方向的梯度算子模板,通过比较8个方向梯度的大小,进一步确定其是否为噪声点;最后对噪声点进行邻域的中值滤波。实验结果表明该算法在滤除噪声的同时,很好地保存了图像的原始信息且有较好的信噪比。

关键词: 椒盐噪声, 方向梯度, 极值, 中值滤波