Research Of Single Image Motion Deblurring And Video Sequence Stabilization

In the process of video sequence acquisition, it will produce image degradation inevitably. For example, there will be blur or unstable in the video sequence when the carrier of video capture device has irregular jitter during the shooting, etc. Image restoration technique refers to the image processing after which we will get clear, stable and high quality video sequence. It occupies an important role in the military, industry, civil and other fields. Therefore, image restoration has always been a focus of concern.For the problems mentioned above, the image restoration technique discussed in this paper includes motion deblurring and digital image stabilization. In motion deblurring field, by reasons of the ill-posed of image restoration, it is not easy to get perfect result when we use inaccurate PSF estimated by early approaches, in the paper, we add a constrains on the recovery equation using a new prior knowledge of image, then describe an idea using multi-scale and edge information of image to acquire PSF as accurate as possible, according to the experimental results, this method can get good recovery result. In digital image stabilization field, firstly, this paper studies a digital image stabilization algorithm based on representative point matching algorithm, we achieve and verify the stabilizing effect of this algorithm, since the algorithm only be used in the case of translational motion, we propose a method to solve the case which contains image rotation. secondly, this paper combines the corner detection and optical flow tracking method while taking advantage of OpenCV library to implement a system of digital image stabilization, the experiments show that this system shows a good result of stabilization of video sequences.