Context-based coherent surface completion

This article presents a novel algorithm for surface completion. We propose a new objective-coherence. We formulate the coherence requirement as a global error minimization process and propose a novel modification technique that minimizes this error. It utilizes our new patch similarity metric, which takes into account both local and global shape attributes. The quality of our completion is demonstrated on challenging cases. These include both general objects and archaeological objects. The latter is especially interesting, not only because of its importance, but also because archaeological objects are usually eroded and noisy and hence challenging. Furthermore, these objects are handmade and thus, though they look symmetric at a first glance, in fact they are inherently asymmetric and no subsurface exists that can be used as-is to complete the hole. In the future, we would like to explore different methods for allowing patch similarity with scaling differences.