An Improved Method of CRIMINISI Algorithm
Tao Yang, Junhong Zhao · 2021
Nowadays, the research on image inpainting technology is a research topic of great significance in computer vision and graphics. As for the missing part of the image, humans can only analyze the remaining areas based on their own visual senses and experience, and then fill in the remaining missing areas because they do not know the original appearance of the image. In general, image restoration is a technology based on human vision and experience. In 2003, Criminisi et al. proposed a patching algorithm based on sample blocks. The algorithm gives priority to pixels based on the confidence level and adopts the best matching module method. This article provides a method to improve its confidence, changing its factor multiplication to linear addition and adding an additional judgment factor, which improves the effect of the inpainting.