versatile video

IEEE Transactions on Circuits and Systems for Video Technology

The latest High Efficiency Video Coding (HEVC) standard has been increasingly\napplied to generate video streams over the Internet. However, HEVC compressed\nvideos may incur severe quality degradation, particularly at low bit-rates.\nThus, it is necessary to enhance the visual quality of HEVC videos at the\ndecoder side. To this end, this paper proposes a Quality Enhancement\nConvolutional Neural Network (QE-CNN) method that does not require any\nmodification of the encoder to achieve quality enhancement for HEVC. In\nparticular, our QE-CNN method learns QE-CNN-I and QE-CNN-P models to reduce the\ndistortion of HEVC I and P frames, respectively. The proposed method differs\nfrom the existing CNN-based quality enhancement approaches, which only handle\nintra-coding distortion and are thus not suitable for P frames. Our\nexperimental results validate that our QE-CNN method is effective in enhancing\nquality for both I and P frames of HEVC videos. To apply our QE-CNN method in\ntime-constrained scenarios, we further propose a Time-constrained Quality\nEnhancement Optimization (TQEO) scheme. Our TQEO scheme controls the\ncomputational time of QE-CNN to meet a target, meanwhile maximizing the quality\nenhancement. Next, the experimental results demonstrate the effectiveness of\nour TQEO scheme from the aspects of time control accuracy and quality\nenhancement under different time constraints. Finally, we design a prototype to\nimplement our TQEO scheme in a real-time scenario.\n

Enhancing Quality for HEVC Compressed Videos