Video inpainting of objects using modified patch based image inpainting algorithm

An efficient moving object completion algorithm suitable for real-time content-based multimedia application system is proposed in this paper. In the proposed method, 3D volume of video is sampled into 2D spatio-temporal slices. Background subtraction technique is used for object detection and tracking. Then, exampler-based image inpainting is applied to complete partially damaged object trajectories in 2D slices. After combination of completed slices, sequences of virtual trajectories are formed to retrieve the most similar postures among the number of postures. Key posture selection and indexing are used to reduce the complexity of posture sequence retrieval.