When transmitted through a poor quality network or stored on an unstable storage media, block-based code images will experience the block loss. To restore damaged images suffering from block loss, Best Neighborhood Matching and Jump and Look-Around BNM provide the most effective image restoration. However, while BNM offers good restoration quality, it requires a large calculation time. By “JUMP” method, JLBNM can effectively shorten the computation time but this comes at the cost of a loss in quality. We have therefore proposed a new image inpainting technique that uses theWavelet Domain to deliver fast computation time and high restoration quality Wavelet Stage BNM. Our proposed reconstruction algorithm includes three optimization techniques change of analytical domain, consideration of texture composition and a new decision-making mechanism: DirectionalWaveletWeighted Method. Theoretical analysis and experimental results demonstrate our method delivered fast computation time and high restoration quality.
淡江理工學刊=Tamkang Journal of Science and Engineering 12(4)，頁439-448