Adaptive video enhancement using neural network

This paper proposes an adaptive video enhancement method using neural network (NN). Inaccurate separation of the composite video signal produces cross-luma artifacts such as dot crawls and blurs edges of an image. Even digital televisions suffer from these artifacts because high definition contents are based on old technology. In order to correct old artifacts we propose two types of NN filtering. Each decoded video frame is segmented into regions of artifacts and enhanced by proposed NN filters. Proposed NN filters adaptively enhance video frame based on the region classification index of a pixel. Experiments using a number of video sequences verify the effectiveness of the proposed algorithm.