A Modified Variational PDE Model For Image Deblurring And It's Applications

Image denoising and deblurring were two of the most important tasks for image pro-cessing. In this study, we first summarized the fast algorithm for convolution, which canbe modeled as image blurring. Then a new e?cient adaptive method――GM-TV modelwas proposed for recovering image from the mixed noisy data. Based on the perspective ofstatistics and variational techniques, we established a new cost function and obtained someEuler-Lagrange PDE equations. By solving these equations which we acquired an estimationof the original clean image. Compared to other models, some new and modified terms inour methods enable us to adaptively adjust the regularization term of the cost function,furthermore, this approach could automatic produce some weights to reduce the in?uence ofthe di?erent noise. All of these could help us restore images which have mixed noise. In thefollowing, we analysed the algorithms which were employed in this paper and had found aconvergence condition. At the same time, some useful conclusions which could guide us howto choose parameters would be shown in this study. According to these theoretical results,we had speeded up the algorithm and improved the quality of restored image. Numbers ofexperiments are performed and corresponding results are shown, which could be seen thenew model was very e?cient. Finally, a parametric blind deblurring model was proposed todeal with the blurred image of registration number of the car and it's usefulness could alsobe shown in some real applications, such as treating with the blurred photos which shootingby the digital camera.