An image inpainting technique based on illumination variation and structure consistency

Image inpainting is that restores the damaged region of interest by user specified or removes areas in an image. For the purpose of patching up automatically, filling in large-sized damaged areas in the image, and lowering down the operation complexity, we design an exemplar-based image inpainting as our main framework. In order to keep the structure consistency between the damaged areas and the non-damaged areas, we redefine two factors, confidence and variation, which use to determine the priority of the filling order of target (damaged) regions. With these two newly defined factors, we propose a dynamic rule to choose the best target region and the suitable patch size in each filling step. To solve the challenging problem of repairing structure in image completion, we introduce a novel image completion approach based on automatic salient structure propagation. Experimental results demonstrate that this approach can make an effective inpainting and visual quality.