Image Restoration using Prioritized Exemplar Inpainting with Automatic Patch Optimization

Image analysis and restoration has been carried out by texture synthesis for large regions and inpainting algorithms for small cracks in images. A new approach that allow simultaneous fill-in of different structures and textures is being proposed in present study. A combination of structure inpainting and patch-based texture synthesis with patch size optimization is carried out (termed as ‘optimized patch based inpainting’) for filling and updating the target region that shows additional advantages over earlier approaches. The algorithm here uses the patch based inpainting having isophote driven patch-based texture synthesis at core. In present algorithm, ones the user selects the target regions, the algorithm automatically search and fills-in these regions with best matching information surrounding it. Patch-based filling is found to improve execution speed and robustness significantly over pixel-based fill. The fill-in is done in such a way that structure information arriving at the region boundaries is propagated inside. A number of examples on real and complex images are considered to demonstrate the effectiveness of the algorithm.