【摘要】 非局域均值滤波器是目前图像处理领域一个重要去噪算法。传统的非局域均值滤波器,主要针对单模态图像进行设计。将其应用于跨模态图像处理时,由于图像间的结构差异性,往往导致其滤波输出产生纹理复制的缺陷。针对这一问题,本文引入相关系数进行图像间的结构相似性刻画,并将其用于滤波器权值的动态正则化约束,最终得到一个适应跨模态场景的非局部均值滤波器算法。实验结果表明了所提算法的有效性。更多还原

【Abstract】 Nonlocal mean filter is an important denoising algorithm in the field of image processing. The traditional nonlocal mean filter is mainly designed for single mode images. When it is applied to cross-modal image processing, the structure difference between images often leads to the defect of texture replication.In this paper, the correlation coefficient is introduced to describe the structural similarity between images and embedded in dynamically-regularizing the weights of the filter. As a result, a novel version of non-local mean filter algorithm is obtained, which is adapt to cross-modal scenarios. Finally, the experimental results show the effectiveness of the proposed algorithm.更多还原