基于超像素分割的非局部均值去噪方法

针对非局部均值(NLM)去噪算法在变化丰富的纹理区域采用平移窗口的方法选择相似块的不足进行了研究，提出一种基于超像素分割的非局部均值去噪算法。该方法充分考虑非局部均值去噪算法中相似性对噪声去除的影响，利用经过超像素分割处理得到的图像块内部相邻像素间以及纹理边缘都具有一定相似性的特点，在超像素分割块基础上优化纹理区域相似窗口的选择策略，提高图像块与中心像素块之间的相似性，从而达到提升非局部均值算法的去噪水平、边缘纹理不被模糊的目的。在多幅经典自然图像上的实验结果表明，该方法能够有效的去除图像中包含的噪声信息，相比于传统的非局部均值方法，保留了更多的纹理信息。

Non-local mean denoising algorithm based on superpixel segmentation

Foucsed on the deficiency that non-local mean (NLM) denoising algorithm is not adapted to the parallel selection of similar window in the texture region ,this paper proposed a non-local mean denosing algorithm based on superpixel segmentation .This algorithm fully considered the influence of similarity in the respect of noise removal while using non-local mean algorithm, took advantage of characteristics of the similarity of neighboring pixels and boundary texture in the image block obtained by super-pixel segmentation.By optimizing the similar window selection strategy based on the super-pixel segmentation block and improving the similarity of the image blocks , so as to enhance the denoising level of the non-local mean algorithm and avoid blurring in the edges. Experimental results on a number of classical images show that the proposed algorithm can effectively remove the noise information contained in the image and retain more texture information compared with the traditional non-local mean method.