Study On Digital Image Inpainting Algorithms Based On TV Model And Snake Model

In this thesis, firstly we give a brief introduction of image restoration, and analyzeadvantages and disadvantages of extant common image inpainting models.Secondly we present an improved adaptive image inpainting algorithm based on TVmodel by considering the relativity between the points in damaged regions and thesurrounding pixels. The method gives a unified expression of different algorithms fordifferent features of the damaged area by introducing different weights. In addition we setdifferent parameters during the iterative process to solve the sensitive problem in TV modeland get better restoration.Lastly we also give an adaptive Snake-TV model in color image inpainting byincorporating the image enhancement term of Snake model into the total variation (TV)model and taking into account the damage area’s shape. In thermal diffusion process, thisalgorithm evolutes curve to boundary to repair the broken edges of fouling image whichkeeps the image visual connectivity. Experimental results show that this model is muchbetter than the original TV model under the same iterations, and better than Snake-CDDmodel with much less time.