结合各向异性扩散和高阶扩散的视频盲去模糊复原方法

A Video Blind Deblurring Method Based on Anisotropic Diffusion and High-order Diffusion

为了能够有效的去除模糊,得到高质量的视频图像,提出了一种结合各向异性扩散和高阶扩散的视频盲去模糊方法.利用基于全变差的各向异性扩散和shock滤波器来准确提取模糊视频中每幅模糊图像帧的大尺度边缘,并利用提取的大尺度边缘,结合稀疏性约束来对模糊退化函数进行准确的估计;在模糊视频图像的复原阶段,为了保证连续的视频图像帧之间的时间一致性,在视频的时间轴上,提出了一种基于高阶扩散的连续性正则化约束项,以在时间上对连续的视频图像帧进行较好的一致性约束.提出的方法能够在较好复原模糊视频的同时有效地抑制复原视频中的振铃瑕疵,同时进一步增强复原视频中的时间一致性.

In order to remove the blur and obtain the high quality restoration video, this paper proposes a blind video deblurring method which combines the anisotropie diffusion and high-order diffusion. This study uses the anisotropic diffusion based on total variation and shock filter to accurately extract the large scale edges of each fuzzy image frame in a fuzzy video, and uses the extracted edges to accurately estimate the fuzzy degradation function. In the restoration stage of fuzzy video image, the study proposes a continuity regularization constraints based on higher-order diffusion in order to guarantee the time consis- tency between continuous video frames base on the time axis of the video to carry out a better consistency constraint of continuous video frames in time. The proposed method can effectively restrain the ringing in video restoration while restoring the fuzzy video and further enhance the temporal consistency in the restored video.