摘要
非局部均值滤波器通过欧氏距离来衡量非局部区域内像素块之间的相似性,取得了较好的去噪效果.但其对局部性考虑不足,易导致一些非周期性的有用细节在图像去噪过程中被光滑掉.针对此问题,引入空域局部、非局部联合自适应方法,对原滤波器进行改进;同时,考虑到多模态图像在实际中的应用愈加广泛,将所设计滤波器推广至跨模态场景,得到了跨模态空域自适应联合均值滤波器.经典图像实验的主观视觉效果与客观的量化指标均表明,所设计的滤波器较原算法取得了更好的滤波性能.
Non-local means filter uses Euclidean distance to measure the similarity of gray values between pixel blocks in a non-local area. In this sense, the filter does not take the locality into account, and thus leads to several local and aperiodic details being over smoothed in denoising process. To address this problem, this work pursues a local and non-local adaptive strategies to improve the performance of non-local means filter. Meanwhile, considering the fact that multi-modal images are widely used in practice, we also focus on extending non-local means filter to a cross-modal version. Finally, a novel filter is achieved which is cross-modal and can adaptively tradeoff between local and non-local implementation. Experimental results show that the performance of the proposed method is more powerful than the original one.
作者
杜婉君
孙忠贵
Du Wanjun;Sun Zhonggui(Ji Xianlin Honors School,Liaocheng University,Liaocheng 252000,China;School of Mathematical Sciences,Liaocheng University,Liaocheng 252000,China)
出处
《南京师范大学学报（工程技术版）》
CAS
2022年第1期52-58,共7页
Journal of Nanjing Normal University(Engineering and Technology Edition)
基金
国家自然科学基金项目(11801249)
山东省自然科学基金项目(ZR2020MF040)
聊城大学开放课题(319312101-01)。
作者简介
通讯作者:孙忠贵,博士,教授,研究方向:图像处理、机器学习.E⁃mail:altlp@163.com。
