Algorithm Research On Restoration Of Image Motion Blur

Blurred image restoration has been a hot research topic in the image processing field, which has broad application prospects. There are many causes leads to blurring image, such as turbulence, defocus and camera shake etc, and the blurring due to camera shake or relative motion of objects is called motion blurring. In daily lives, reduction shake blur on camera, video surveillance recovery and motion license plate recognition all requires such motion blur restoration technology. Motion blur restoration has developed to lots of directions, and relative algorithms are endless, but the thrust is always to eliminate blur, while improving image clarity, suppressing ringing, and retaining more details.In this paper, the motion blur restoration is divided into two phases, blur kernel estimation and clear image restoration. The main algorithm models and blur kernel estimation methods of image deblurring are deeply analyzed, and the regularized restoration methods based on prior constraints are simultaneously explored. Starting from the study of motion blur model, the related work of this paper mainly includes the following aspects:1) The development status of image motion deblurring at this stage is outlined and the difficulties of image restoration are analyzed. Then the main image deblurring models and prior constraints are introduced, and the quality evaluation criterion of image restoration are discussed.2) The accurate estimation of blur kernel for uniform linear model is analyzed and implemented. For RL restoration algorithm’s strong ringing effect, method based on the gradient of attenuation is used to suppress the ringing effectively.3) Focusing on the fast kernel estimation algorithm based on prediction on the fuzzy edge information of the image, strategies in optimization and modification of blur kernel are utilized to improve accuracy of the blur kernel estimation. Then the gradient ultra-Laplace distribution constraints is introduced to improve the quality of the restored image.4) The restoration algorithm based on L1 norm constraint restoration, solved by the split Bregman, is utilized to improve the quality of the restored image after analyzing four algorithms based on sparse constraint by gradient.5) After analyzing local and non-local recovery characteristics of image, the guide filter algorithm is introduced. Combined with the fast and efficient split Bregman method, the restoration algorithm joint local and non-local information by guide filter is proposed. Enhancing the recovery effect, de-noising and preserving more details are achieved in blind motion blurred image restoration.By contrasting the proposed algorithms with the original algorithms through experiments, the proposed algorithms are proved to achieve better results in the suppression of ringing and blur kernel estimation, and constraint restoration of image.