Decoupling: Image occlusion removal based on image segmentation and image inpainting

Abstract

Abstract In daily photography, there are often occlusions in the picture, which affect the overall image perception, so we hope to achieve object removal through image processing. At present, this research field is relatively new, has not yet formed a completed implementation system, and most of the implementation effect is rough, this study plans to achieve object removal by combining semantic segmentation and image inpainting. This paper also compares our model qualitatively or quantitatively with existing methods and conduct ablation experiments to fully illustrate our ideas. Extensive experiments show that the method obtains fairly good object removal performance.