Abstract
We propose an image inpainting algorithm based on adaptive edge-preserving propagation for structure repairing. Neighbouring information is progressively propagated into damaged region. The optimal size and location of the window containing damaged pixel are adaptively chosen according to the intact degree and colour distribution. To preserve sharpness of edges, contributing weights of the pixels in neighbouring window are decided by their direction with isophote and distance with damaged pixels. Compared with typical partial differential equation (PDE)-based methods, the proposed approach is more concise and efficient, and can give satisfactory results for structural information repairing. Experiments are carried out to show effectiveness of the method.
This work was supported by the Natural Science Foundation of China (nos. 60872116, 60832010 and 60773079), the Shanghai Rising-Star Program (no. 10QH14011), the Key Scientific Research Project of Shanghai Education Committee (no. 10ZZ59), the Shanghai Specialized Research Foundation for Excellent Young Teacher in University (no. slg09005) and the OECE Innovation Foundation of USST (no. GDCX-Y-103).