Domain-based structure-aware image inpainting

Although image inpainting has been extensively studied in recent years, some problems in this area are still open. In particular, the structure restoration is one of the difficulties due to the incompleteness of the reconstructed structural information. The less reasonable filling order and the ignorance of local consistency of the image would also easily lead to undesired repairing results. To remedy the above problems, this paper proposed a new domain-based structural-aware image inpainting method. We specially designed a new iterative structure searching algorithm which can restore more complete and reliable structural information. The adjacent patches were connected to form a repairing domain which serves as the minimal repair unit. The introduction of the domain ensures the coherency and searching accuracy of the repairing results. Moreover, we introduced a novel repair order calculation method which can greatly reduce the influence of the error propagation in conventional methods. Various experiment results demonstrated the effectiveness of our method.