An Image Inpainting Algorithm Based On Texture Synthesis

Digital image inpainting mainly aims at filling in missing pixels in an unknown region of an image in a visually plausible way, it is an important branch in the field of digital image processing. And it is also one of the important research fields in computer graphics,computer vision and image processing. Currently, the primary methods of image inpainting include the algorithms based on Partial Differential Equations(PDEs) applies to the recovery of scratches, blots and characters. However, the "Texture Synthesis" algorithms obtains good effects when it comes to the missing region with large scale.In this paper, we studied several typical image inpainting models based on the basic knowledge of this technique; Generally we use the model based on the variational functionals\PDE to repair the image whose damaged area is relatively small. There are three typical repair models. (1) BSCB model.In this model the information generated by PDE equation slowly extends to the area to repair the damage, then gradually complete the painting(2)TV model. This model is also based on a mathematical model PDE equation, and because the extension information is anisotropy, its main advantage is the ability to maintain the edge. The effect is better than BSCB model. (3) the CDD model.CDD model introduced scalar curvature of characterize the geometry information Isophote line to conductivity in traditional TV model, making the spread of the large curvature becomes strong,and the spread of the small curvature disappearing. This model make up the shortcoming that TV inpainting model sometimes can not guarantee connectivity,and the model can connect small edge damage eara better.Then this paper describes the general image inpainting algorithm which is based on texture synthesis. This inpainting algorithm is suitble for the large damaged area of the image. This paper studies several inpainting method based on texture synthesis, focusing on the study of Criminisi inpainting algorithm based on sample,and this paper presents a new image inpainting algorithm based on texture synthesis. This algorithm keeps the edge feature of injured image by dealing with the priority of the boundary pixels and the weight of the pixels in neighborhood. Besides this, the texture details and structure information are fully considered. And the same time, blurring effect is avoided. This algorithm can also achieve fairly good results when do the big damaged blocks inpainting. In addition, this algorithm improve the difficult of the exemplar-based inpainting method searching the source patch globally easily leads to some matching errors, and such errors will accumulate and propagate into other areas quickly, even making the results completely unacceptable. When we find the maching block,we combined with gradient information and histogram information to find the best matching block. Thereby reducing the mismatch phenomenon generated image during restoration.,and in the edge of the continuous error inpainting information is also rare.In this paper, some algorithms are successfully tested and proved the feasibility by the software MATLAB.