Blind Restoration Of A Single Motion-blurred Image Based On Multi-channel Constraints

Images with uneven quality,especially blurred images,will bring a lot of inconvenience to our life and scientific research.Therefore,blind image restoration is an important research topic.Among them,which kind of blind restoration algorithm should be used is one of the most important aspects of current research.In this paper,the classical variational energy equation is chosen to realize the restoration of fuzzy images.The variational energy equation is mainly composed of two items,namely data items and rule items.The major algorithms usually choose TVL1 and TVL2 as data items,but the selection of rule items is generally the key to solving the problem.In this paper,we select the normalized nonlinear diffusion term,the normalized curvature term and the L0 constraint term as rule terms.The selection of these rule items solves the image restoration problem well,retains the main features and details of the image,and the restored image is smoother,and the effect of maintaining the edge and enhancing the edge is better.Moreover,the energy functional convergence is fast,and the image recovery speed and the speed of obtaining a clear kernel function are also faster.Data standardization is to ensure that the energy trend is optimal when solving energy equation,that is to say,it shows a downward trend.In order to simplify the solution and improve the solution rate,the fast algorithm ADMM and the split Bregman algorithm are introduced to solve the problem.The effectiveness of the proposed algorithm is verified by experiments.The currently proposed MTV-based models or classic TV-based models generally have some problems such as low computational efficiency and poor recovery of recovery image details.However,the proposed algorithm not only has better visual evaluation and objective evaluation,but also has higher computing rate.In fact,the classic TV(Total Variation)model has long been applied to image deblurring,but since the color image is a multi-layer coupled image,the classic model can not solve the image deblur problem very well,so we are proposing more effective rule items,at the same time,a multi-channel approach was introduced.We first perform graphics processing on each single-layer channel of the color image,and then combine the processing results of each layer to complete the final restoration effect map,which not only can deblur the color image,but also collect various useful information of the layer,in the end,can achieve a more detailed and clearer restored image than the image restoration of the grayscale image.