Application Research On Image Deraining And Desnowing Based On Matrix Completion

Video and image,as the main form of visual information,are the main channels for people to obtain information.When obtaining visual information in rainy and snowy weather,rain and snow will reduce the visual effect of video image,it causes information to be occluded or missing,which makes subsequent analysis and processing more difficult.Therefore,removing the influence of bad weather can effectively improve the visual effect of video image and increase the accuracy and reliability of video image analysis and processing.In this paper,we focus on the deraining and desnowing problem of video image,and carry out in-depth research and analysis on the existing technology of video image deraining and desnowing.Based on the classic Kim method,this paper proposed an improved deraining and desnowing method.Since the Kim method does not fully consider the parameter adaptation problem,this paper introduces an adaptive parameter to the second item of the classic Kim model,which is for improving the effect of video image deraining and desnowing.Then combined with the noise benefit idea of the noisy expectation maximization algorithm,rain and snow are removed like noise in video image.The model with adaptive parameters is abstracted into two parts:background and noise,and based on that to propose an adaptive parameter model on matrix completion in this paper.Based on the improved rain streak removal model,this paper proposes an adaptive deraining and desnowing algorithm based on matrix completion.Inspired by the EM algorithm and the noisy expectation maximization algorithm,the parameterized noise expectation maximization theorem is defined in this paper.And the theoretical proof of the proposed algorithm is demonstrated.Firstly,the synthetic rain videos are used to verify the validity of the parameter in the experiment.Secondly,this paper applies the grid search method to finding the bestremoval parameters.Finally,this paper verifies the effectiveness and excellence of the proposed algorithm by various rain and snow videos.In order to further verify the deraining and desnowing performance of the proposed algorithm,it is compared with the classic video deraining and desnowing methods.The six image quality assessment indicators are used to evaluate the deraining performance of the proposed algorithm on various types of rain and snow video.The experimental data show that the proposed algorithm can effectively detect and remove rain and snow in video image,and retain the information and scene structure in video image.The restored video and image on algorithm of the paper,are closer to the real video and image,and its deraining performance is better than that of comparison method.