Removal of Occlusion for Abstraction of Class - Room Videos

In this paper, we examine the case of occlusion removal from a class-room video in a remote class application. Written text on writing-board in sequence of video frames is occluded by a human subject. Our goal is to remove the occlusion, in our case it is human subject. As the human subject moves across the writing-board, he or she will occlude different parts or regions of the writing-board. The region of occlusion can be replaced by the contents of the frame in which the same region not occluded by the human subject. Thus, it requires to segment each of the frames into two types of region namely occluded and unoccluded. The frames are segmented based on a supervised statistical analysis. For decreasing the computation efficiency of proposed segmentation method, a multi-resolution analysis based on haar wavelet is implemented. In order to fill gaps in the output of the segmentation method, certain morphological operation are implemented which gives the final segmented regions in each frame. Finally, the human occlusion is removed based on the segmentation results by appropriate replacement of the content from the other frames.