Image Inpainting Through Quality Based Patch Selection Matrix

Image inpainting refers to the set of techniques which include filling-in of missing area (known as holes) in a picture such that modifications made to the picture are unnoticeable. The algorithms for image inpainting that are proposed in the literature are rooted on the concept to fill-in the holes by means of available information in the surroundings. This information can be automatically detected or hinted by the user. This paper propose a modified Criminisi’s exempler based image-inpainting method in which quality based patch selection is performed in the process of inpainting of digital images to recover the lost part of the image in a visually plausible way such that the changes made to the image are not detected by the normal user. Here, a measure of SSIM (Structural Similarity Index Metrics) is integrated in the process of inpainting, where, confidence term, data term and priorities for the patches are based on SSIM index that compares local patterns of pixel intensities that have been normalized for luminance and contrast. After the image-inpainting is completed using this proposed modified criminisi’s inpainting algorithm, a single-image superresolution (SR) approach is employed to enhance the resolution of the inpainted image consequently improving the quality of the resulting image. IndexTerms Image-inpainting, Image-processing, texture-synthesis, super-resolution, ssim ________________________________________________________________________________________________________