New Interpolation Algorithm for Image Inpainting

In this paper, the relationship between image pixels’ coordinators and their colors is considered as a mapping function. A novel algorithm based on radial basis function network is applied to construct a local best approximation of this function, and then the damaged pixels are restored via interpolation. In the algorithm, the mapping function is constructed pixel by pixel to minimize the interpolation error and decrease its complexity. To proceed over lager damaged areas, we compute interpolations of different overlapping coefficients to combine global frequency and local spatial information together. This contributes to restore more detail information. Experimental results show that the algorithm can restore the missing information correctly and can be effectively used in reconstruction of complex signal functions.