Improving Exemplar-based Image Completion methods using Selecting the Optimal Patch

Document Type : Research Paper


Iran University of Science and Technology, Tehran, Iran


Image completion is one of the subjects in image and video processing which deals with restoration of and filling in damaged regions of images using correct regions. Exemplar-based image completion methods give more pleasant results than pixel-based approaches. In this paper, a new algorithm is proposed to find the most suitable patch in order to fill in the damaged parts. This patch selection method is recommended for exemplar-based approaches. The proposed method consists of three modifications: First, the selected patch’s size is considered bigger and then different weights are applied to different sections of the patch. Second, harmonic inpainting method is used in order to reduce the error propagation. Third, limiting the search space is done to improve and reduce the executing time. The acquired results depict that this method is effective and there is an improvement for quality of exemplar-based approaches.