Image Inpainting Based on Wavelet Decomposition
Hongying Zhang, Shimei Dai · Procedia Engineering · 2012
Image inpainting is an important problem image processing. It is a difficult problem to simultaneously fill-in the texture and structure in regions of missing image information. In order to inpaint the damaged image with both missing the structure and texture information, an image inpainting algorithm based on wavelet decomposition is presented. First the damaged image is decomposed into structure sub-image and texture sub-image using the wavelet transformation. Then, the sub-image with the region of missing information in the structure is reconstructed by Curvature-Driven Diffusions (CDD) algorithm, while the same region in the texture sub-image is filled-in with the improved texture synthesis based on exemplar; Finally, the restored image is given by recombining the structure and texture restored results. A large number of experiments show that the proposed algorithm can quickly and efficiently restore the structure and texture information at the same time, and the visual effects and the Peak Signal to Noise Ratio (PSNR) is better than the similar algorithms.