Image Colorization And Restoration Method Based On Total Variation And Deep Prior

Image colorization is to add color to a gray image in a suitable way to enhance the visual effect of the image.Image restoration refers to restore the true appearance of degraded image by using the prior knowledge of image quality degradation process.Image colorization and restoration are the premise and foundation of high-quality image processing and analysis.This paper studies image colorization and restoration method based on total variation(TV)and deep prior.The main research contents and innovations are as follows:(1)Combining total variation to locate color boundary and convolutional neural network(CNN)to learn image detail features,a TV model with deep image prior is proposed for image colorization under Plug-and-Play(Pn P)framework.Furthermore,the corresponding numerical algorithm of the proposed model is designed by using the alternating direction method of multipliers(ADMM).Numerical experimental results verify the effectiveness of the proposed model and algorithm.(2)Using the framework of Regularization by Denoising(RED),a TV model with deep image prior under RED framework is proposed for image colorization.Furthermore,the corresponding numerical algorithm of the proposed model is designed by using the ADMM method.Numerical experimental results show that the algorithm not only improves the operation speed,but also has better colorization effect.(3)For large-area high-brightness fuzzy images,a blind image restoration model is proposed based on NN-α regularization and deep fuzzy kernel prior.Furthermore,the numerical algorithm for solving the model is designed by using the alternating minimization.Experimental results are reported to show that the model can estimate the fuzzy kernel accurately and restore the large area of the high-brightness fuzzy image effectively.