부산대학교 도서관

To remove objects from digital images, certain measures are required to restore the empty and damaged regions in an image. The process of restoring a corrupted image by filling empty regions with background textures is called image inpainting. This study proposes a new algorithm to suppress the generation of texture garbage in the image inpainting process. We analyze the characteristics of the projections onto convex sets (POCS)-based and static patch-based exemplar search techniques and propose an adaptive algorithm that can compensate for the shortcomings of both methods. The experiments showed that the proposed method naturally inpainted empty areas while effectively suppressing the generation of texture garbage.