Improved Criminisi Algorithm Based on a New Priority Function with the Gray Entropy
Xiang-yan Xi, Fulong Wang, Yefei Liu · 2013
In the image inpainting process, the data term of the Criminisi algorithm depends on the shape of the manually selected target region and the confidence drops to zero rapidly, resulting in in painting sequence deviation which finally influence the in paint effect. Then we introduce the entropy to improve the data term, in this way another priority function will be defined. Experiments confirm that the improved algorithm can eliminate the dependence on the shape of the target region and the confidence will not drop to zero rapidly again. Experiments show that the algorithm will repair the image with pure texture, strong edges and purely synthetic images better.