基于多尺度相似先验的非局部图像去噪算法

图像去噪是图像处理领域的一个基础研究课题,利用正则化建模方式解决图像去噪问题的关键在于正则化约束项的选择。通过分析图像结构信息,文章假定图像存在多尺度的结构特征,提出了以多尺度相似先验作为正则化约束项的非局部图像去噪模型。该算法利用奇异值分解和硬阈值方法对获得的多尺度相似矩阵进行协同去噪,通过数值实验表明,可以获得性能较好的去噪效果。

Non-local Image Denoising Algorithm Based on Multi-scale Similarity Prior

Image denoising is a fundamental problem of image processing field.This paper employs the regularization method to address the problem of image denoising.The performance of the regularization method mainly depends on the image priors involved.By analyz?ing the image structure information,this paper assumes images have multi-scale structure features.Then propose the model which re?gards the multi-scale similarity as image prior to handle non-local image denoising problem.The algorithm uses SVD and hard thresh?old technology to denoise the multi-scale similar matrix.Extensive experiments demonstrate that the algorithm can exhibit better denois?ing performance.