An Improved Image Inpainting Algorithm Based on Image Segmentation

Abstract

At present, the image restoration algorithm based on sample block is not ideal for image processing with abundant texture information, and it is easy to produce false matching texture blocks in the process of repairing. An improved image inpainting algorithm based on image segmentation is proposed in this paper to address this problem. Firstly, the watershed image segmentation algorithm is used for image segmentation, which is combined with curvature feature of isophotes to repair the image. Curvature features reflecting the details of texture character are introduced in the data items to calculate the priority. Secondly, the matching of the pixel block must be searched in the image segmentation region, which can be a better solution to avoid the phenomenon of excessive extension of the texture block in the process of image restoration. Compared with the original Criminisi algorithm and other improved algorithms, the experimental result shows that the method we proposed have a larger PSNR value of image, which means to have better repair effects on the image including abundant texture information.