基于分裂Bregman算法的单帧运动图像盲去模糊

提出一种基于分裂Bregman算法的单帧运动图像盲去模糊方法,该算法分为模糊核估计和图像复原两个阶段.在估计模糊核时,首先利用双边滤波器去除图像中的噪声,再采用改进的冲激滤波器对图像进行边缘增强,选取有用的边缘信息估计模糊核,并对估计出的模糊核进行修正,从而得到高质量的模糊核.图像复原阶段,利用分裂Bregman算法交替迭代得到去模糊后的图像.该算法具有降噪和边缘增强的功能,并能保持图像总变分不变,使图像复原效果更好且计算时间有大幅降低.

Single Image Blind Motion Deblurring with Split Bregman Method

A single image motion deblurring algorithm based on split Bregman method is proposed in this paper. It is divided into two stages, blur kernel estimation and image restoration. When estimating blur kernel, bilateral filter is used to do image denoising, and then improved shock filter is chosen for edge enhancement. After selecting useful edges and fine-tuning, a high-quality blur kernel is obtained. When image restoration, a clean image can be obtained using iterated split Bregman approach. This algorithm has the effect of image denosing and edge enhancement, at the same time of keeping total-variation unchanged. Experimental results show that this algorithm has better visual effect than other exiting algorithms, and computation time is greatly reduced.