一种去除椒盐噪声的自适应非局部均值滤波

An Adaptive Non-local Mean Filter for Salt Pepper Noise Removal

非局部均值利用图像的非局部相似信息实现图像噪声的消除,但未考虑到噪声之间的差异性与特殊性,其去除噪声与修复图像详细信息能力有限。为了有效去除图像的椒盐噪声,提出一种充分利用椒盐噪声特性与图像的非局部相似信息的自适应非局部均值滤波算法。该算法通过自适应均值实现对椒盐噪声点的检测,进而采用与图像局部方向一致的长方形相似窗口进行非局部均值滤波与信息的重建。对比实验表明,提出的自适应非局部均值滤波算法有效地利用椒盐噪声的特性与图像的自相似信息,在消除椒盐噪声的同时,更好地重构图像的结构信息,峰值信噪比与结构相似度高于目前其它非局部均值滤波,图像去噪性能更优。

The non-local mean method for removing noise focus on image non-local self-similarity,without considering the differences and particularity between the noises,thus resulting in limited performances of removing noise and restoring detail information of image.In order to effectively remove the pepper salt noise,an adaptive non-local mean filtering algorithm is proposed which makes full use of the pepper salt noise characteristics and the non-local similar information of the image.The algorithm detects salt pepper noise areas by adaptive mean filtering,and then adaptive non-local mean filtering with directional rectangular patches is used to remove noise and reconstruct the areas.The comparing experiment results show that the proposed adaptive non-local mean algorithm can effectively use the properties of the salt pepper noise characteristics and image self-similarity information.It can remove salt pepper noise effectively and reconstruct the image information better than other local-based methods at the same time,with higher peak signal to noise ratio and structural similarity value,showing good image denoising performance of the proposed method.