A Inpainting Model Using The Total Variation And Wavelet Method

Image inpainting is an important problem in image processing. The main goal is to restore the damaged image caused by cracks or scratchs, and to remove the objects or disocclusion in computer vision and error conceal in videos. In recent years, reachers have developed several methods for image inpainting. However in demestic, this area is just at development stage.This thesis introduces the background and significance of image inpainting techniques and reviews the research state both at abroad and demestic. Some applications of image inpainting also can be found. And then we demenstrates the partial differential equation methods in details and summarizes the advantages of this kind of models.Also the models based on texture synthesis in image inpainting are discussed.The thesis put the focus on wavelet image inpainting techniques. A new variational inpainting model is obtained by replacing the TV norm into the H ^-1norm of wavelet coefficients in the TV inpainting model. Experimental results show that better inpainting qualities can be achieved in our new model.