Image Inpainting Based On Different View Images

In this paper, we introduce the problem of image inpainting based on large displacement view images and its solution, including inpaint planar image or approximate planar image or the damaged area of 3-D boject, which is the problem of filling in the occluded or damaged regions of an image using the visible information from other large displacement view images. The key problems are how to convert the visible information in large displacement view images to be useful and how to use converted useful information to repair the target image.For planar image or approximate planar image , we solve it by first dividing all images into different scene planar regions and image matching to transform all image regions into current view thus the visible information can be directly used, then a new inpainting algorithm based on texture synthesis and image fusion or getting a line with the least energy in the target image by using GraphCut to fill the holes. In addition, we solve the ghost effects between the repaired region and the target image by Poisson image fusion to get seamless effect.For the damaged area of 3-D object inpainting in the target image , first we should dividing the 3-D object and the remain information in the target image. Getting the interpolate images of 3-D object and the remain information through several candidate images. Then using the interpolate images to inpaint the target image , the next process of inpainting is similar to above.The experiments show good results for the repaired region through our algorithm.