Combining Texture Synthesis & Inpainting for unwanted object removal and image completion

Abstract – In this paper, a novel algorithm is proposed for removal of unwanted objects from an image and image reconstruction in such a way that alteration in image is undetectable to human eye. Application of this technique includes restoration of damaged photographs and films, removal of superimposed text and removal/replacement of selected objects etc. After user selects the regions to be removed or restored, the algorithm automatically fills these regions with data sampled from reminder of the image. The proposed algorithm combines the strength of “Inpainting” and “Texture Synthesis” together to restore the Structure and texture information of the damaged image efficiently. Computational efficiency of algorithm is achieved by block based sampling process. The algorithm works well on real and synthetic images and is capable of removing large objects as well as thin scratches.