Research On Image Restoration Algorithm Based On DCT Similar Sample Features

Image inpainting refers to the known image features to repair lost or deteriorated parts of the image data,which is an important research direction in the field of digital image processing.The goal of image inpainting is to make the reconstructed image more visually pleasing and difficult to notice its corrupted parts.The traditional image inpainting algorithm generally applies the Delaunay Spline Interpolation(DSI)to initialize the randomly missing image.However,this method may produce improper image structure information.This thesis intends to try out a new image inpainting model based on ?1 norm and a study on DCT similar example-based image inpainting algorithm.The main work is as follows.Chapter 3 proposes a study on DCT similar example-based image inpainting algorithm.This chapter uses the ?1 norm to change the regular items in the original image inpainting model.And then the alternating direction method of multipliers(ADMM)is used to solve this model.There is the convergence proof process of ADMM method in theory.In the model implementation processing,the smooth patch ordering method based on the image patch is applied to initialize the randomly missing image.The experimental results show that the new model can achieve better results.Chapter 4 proposes a new image inpainting model based on ?1 norm.The split bregman iteration algorithm is replaced by the alternating direction method of multipliers in order to solve the image inpainting model in this chapter.Moreover,an approximation method which is employed to simply the image inpainting model.In the model implementation process,the smooth patch ordering method based on the image patch is applied to initialize the randomly missing image.And the discrete cosine transforms tight frame is used to obtain the image local features.Furthermore,the similarity image patch sets can be got.Experimental results show that our algorithm is superior to some state-of-the-art inpainting methods.