The Research And Implementation Of Still Nontextural Images Inpainting Based On GMRF Model

As an important part of image restore technology, digital image inpainting technology means local object (just as letter,object and so on) in picture is filled or removed.In this paper, image inpainting algorithm was deeply researched and discussed from the view of GMRF model on which two new kinds of improvement image inpainting algorithm were based. One is based on GMRF model associate with edge information. The order of inpainting the boundary pixels was determined by priority check. To take the edge features into account, the experimentations indicated that the method could achieve a good consequence. The other is based on multi-scale GMRF model. The approximation of inpainting regions between the whole and the part were finished through the fusing of whole information and part information.The result of numerical experiment show that GMRF model associate with edge information fully consider the edge features, and better maintain the information of edge in damaged regions. The inpainting result made some improved. But, the multi-scale GMRF model could achieve fairly good results. Because whole information of damaged regions was approximated, furthermore, part details restored well.