Image inpainting algorithm based on regional segmentation and adaptive window exemplar
Yang Liu, Tian Xiao-jian, Qing Wang, Shao Shang-xin, Sun Xiao-lin · 2010
This paper presents a novel and efficient algorithm for preventing the block wrongly matching in image inpainting. The proposed technique separates the inpainting process into two different and important steps. First, segment the pre-selected source area, which is called the source region, from the input image. This step can prevent the block wrongly matching effectively. Besides, the adjustment rule of adaptive window is utilized to determine the fixed window size, which could be adjusted according to the image itself, after that search a best-exemplar from pre-selected region to fill the image. This step can improve the quality of image inpainting. The experimental results show the high efficiency and quality of the algorithm.