Intelligent video inpainting system for texture reconstruction

Video inpainting is one of the very challenging problems in computer vision. This approach may be used for removing an undesired object from an input video. In this Paper the undesired object will be removed and the texture is reconstructed in the entire video while maintaining the consistent flow of video. The technique implemented here is for the videos recorded with stationary camera. It involves the steps of preprocessing, motion inpainting, and background filling-in. The step of preprocessing involves segmenting each frame of a video into foreground and background, detecting the foreground objects and then tracking those detected objects. The motion inpainting step will fill the holes created by the occluding object by copying information from object templates generated in preprocessing step. The last step of background filling-in will fill the remaining holes by copying corresponding background from background model. In this paper, the challenges of occluding object removal, occluded object removal, and addressing complete occlusion are handled.