Image inpainting considering symmetric patterns

This paper proposes a novel image inpainting method to remove undesired objects in an image. Conventionally, missing regions are filled in using similar textures in an image as exemplars. However, unnatural textures are often generated due to the paucity of available samples. In this study, we take into account symmetric transformation of texture patterns to increase exemplars. To generate plausible textures in missing regions with variously transformed patterns, we employ two approaches: (1) we use spatial coherence of texture patterns when searching for similar patterns, and (2) we define a new degree of confidence of exemplars for determining pixel values. The effectiveness of the proposed method is demonstrated by comparing results by three methods.