基于偏微分方程的组合视频去噪方法

提出1个基于偏微分方程的组合视频去噪方法.该方法综合利用了变分运动估计方法的精确性和各向异性扩散滤波方法的保边性能等优点.方法的实施分为3步:首先,采用2维各向异性扩散滤波对噪声视频进行初始的去噪.然后,采用Brox等提出的变分光流计算方法对前1步去噪后的视频进行运动估计,目的是计算各帧的前向及后向运动场.最后,采用3维各向异性扩散滤波对运动补偿后的原始噪声视频实施最终的去噪.实验结果表明本文的方法优于经典的各向异性扩散滤波方法.

A Composed Video Denoising Method Based on Partial Differential Equation

Based on partial differential equation this paper present a composed video denoising method.It combines the accuracy of the variational motion estimation methods with the edge-preserving qualities of anisotropic diffusion filters.The proposed algorithm proceeds in three steps:First,a 2-D anisotropic diffusion filter is adopted to initially remove noises from the noisy video.Then the variational optic flow technique of Brox et al.is applied to the previous denoised video,in order to compute the forward and backward motion fields.Finally,a 3-D anisotropic diffusion filter is used to finally remove noises from the motion-compensated noisy video.Experimental results show the proposed algorithm outperforms classical anisotropic diffusion filters.