Research On Image Impainting And Completion

Image inpainting and image completion is a technique to fill the missing or damaged region of an image, using the source region of the image, so that the result is similar to or almost the same as the original image. Image inpainting mainly handles the problem of small region missing, while image completion mainly deals with large region problems. Digital image inpainting and completion can be applied to artwork restoration, image editing (e.g filling gaps, object removal), film postproduction, computer graphics application (e.g texture synthesis), etc.Based on two present image completion algorithms, we propose two modified algorithms, which are our main work of the paper:First, we present a spatial correlation based image completion method. A local search instead of global search, is used to find the best-match patch, which both reduces the search complexity extensively and maintains good completion results. Experimental results are presented, which demonstrate the effectiveness of the algorithm.Second, a parameter adaptive algorithm is presented. To avoid the work of inputting parameter in a heuristically way, we try to get the parameters automatically. Given an initial parameter, the algorithm computes the number of active labels and judges the input parameter with the number. If the parameter is not proper, the algorithm adapts it automatically. In the end, a best parameter is obtained.Finally, we summarize the work in the paper and give some of our ideas about the further development in this field.