A digital image inpainting method based on multiscale markov random field

A novel approach to image inpainting based on multiscale markov random field is introduced. The novelty lies in the combination of the local information via the inner-scale MRF technique and global information via the inter-scale markovian chain to improve the image inpainting quality. The proposed algorithm consists of three steps. In the first step, the image to be inpainted is decomposed into multiple scales. In the second step, the image at the coarsest scale is inpainted based on the MRF model. Finally, the inpainting result is obtained according to the innerscale and interscale markovian relation from the coarsest resolution to the finest one. The experimental results show that the proposal inpainting algorithm achieves fairly good results. ©2010 IEEE.