Research On Image Deblur Algorithms

The objective of image deblurring is to recover the clear image from the degraded image.In practice,since the blurring processing is unknown,blur kernel estimation with high accuracy becomes the key to restore the clear image.The main contributions of this thesis are to obtain a more accurate blur kernel,and then to apply the efficient optimization models to restore the clear image.To obtain accurate blur kernel function,a method is developed to estimate the blur kernel using the RGB channels and then a robust image deblurring model is presented to recover the clear image.The proposed method is different from the existing methods because they estimate the blur kernel only in the gray-scale domain.Since each channel experiences different effects from blur process,by using that information,the estimated kernel should be more accurate.In the development of the proposed method,the significant image edge information is also used as the priori constraint of regularization under the multi-scale condition.After obtaining the blur kernel under the RGB channels,by considering the estimation error in blur kernel,a robust image restoration model is developed based on the framelet system to alleviate the error.The simulation results demonstrate the accuracy of the blur kernel obtained under the RGB channels.Compared with the blur kernel estimated only in the gray domain,the performance of the proposed approach in image deblurring is superior.In the developments of image deblur algorithms,it is noticed that sparse prior plays a very important role in the final recovery performance.To find a more sparse representation for images,the concept of group sparse is utilized.Compared with the traditional sparse representation based on the image block,the similar image blocks are collected to form the group.In doing so,the same image blocks in each group use the same dictionary,and the resulting sparse coding coefficients are more accurate.Based on the group sparse,the blur kernel is also estimated by utilizing the RGB channels and to solve the optimization problem,the split Bregman iteration(SBI)is utilized to efficiently obtain the solution.Simulation results show that the performance by the proposed approach is superior to the state-of-the-art algorithms in terms of both the visual inspections and objective evaluations.