Image Inpainting Based on Image Structure and Texture Decomposition

Image inpainting is a major research content in the field of computer vision and is intended to solve the problem of high-quality repairing of the damaged images automatically. In this paper, considering the image containing structural component and texture component generally, an Adaptive Local Laplacian Filter (ALLF) is first proposed to decompose the image, and the different components obtained through ALLF decomposition are further repaired by different methods. The restoration results are added together to achieve the completion of the whole image. Experiments on synthetic and natural images show the advantages of our approach, especially, in small scratch repair and text removal task.