Image Completion Considering Local Orientations of Rotated Patterns

Image completion yields whole images by producing plausible parts missing due to the removal of foreground or background elements. Conventionally, missing parts are produced by optimizing the objective function, defined based on pattern similarity between the missing region and the remaining image (data region). The resulting image may be compromised, however, by data region pattern variations. Augmenting data region pattern variations positively produced good results, but tends to cause processing search time to mushroom proportionately. To avoid this, we propose pattern extension based on rotating data region pattern variations and minimizing calculation time using the local orientation of rotated patterns. The effectiveness of this approach was demonstrated by comparing conventional and proposed methods.