摘要
随着现代成像技术的不断完善,在一个成像设备上几乎瞬间就可获得多帧图像。针对这些图像在收集、传输和处理中往往会受到噪声污染的问题,文章提出一种新的多重余量Wasserstein正则的多帧图像去噪算法。首先,基于图像总变差约束和余量概率分布之间Wasserstein距离最小化思想,提出多帧图像去噪的多重余量Wasserstein正则模型;其次,呈现直方图匹配运算和Chambolle对偶投影的交替迭代求解算法;最后,展示提出方法对应的数值实验。实验结果显示,提出的多帧图像去噪算法在客观评价和视觉效果两方面都明显优于经典的图像去噪方法。
With the continuous improvement of modern imaging technology, multi-frame images can be obtained almost instantly on one imaging device.Aiming at the problem that these images are often polluted by noise during collection, transmission and processing, this paper proposes a new multi-frame image denoising algorithm based on Wasserstein regularization with multiple residuals.First, based on the constraint of image total variation and the idea of minimizing Wasserstein distance between residual probability distributions, a Wasserstein regularization model with multiple residuals for multi-frame image denoising is proposed.Then, the alternate iterative solution algorithm of histogram matching operation and Chambolle dual projection is presented.Finally, the numerical experiments corresponding to the proposed method are presented.Experimental results show that the proposed multi-frame image denoising algorithm is significantly better than the classical image denoising methods in both objective evaluation and visual effect.
作者
何瑞强
兰旺森
HE Ruiqiang;LAN Wangsen(Department of Mathematics,Xinzhou Teachers University,Xinzhou 034000,China)
出处
《忻州师范学院学报》
2022年第2期1-8,共8页
Journal of Xinzhou Teachers University
基金
山西省应用基础研究青年科技研究基金(201801D221033)
忻州师范学院教学改革项目(JGYB202024)。
