An Adaptive Selection Algorithm of Screen Content Coding Tools for Educational Video System

Screen content coding (SCC) has a high compression performance for screen content. On the other hand, the computational complexity is enormous compared to traditional coding tools. In this paper, we hypothesize that SCC can be used efficiently by pre-determining image features using machine learning. Our proposed method continues to automatically select a highly efficient encoding method without being affected by real-time switching of the screen contents. The proposed method achieved a maximum 70% BD-Rate improvement over traditional VVC encoding and a maximum 38% reduction in computing time compared to using SCC for all frames. In the future, the proposed architecture can be implemented in a hardware encoder to enable fast and efficient video coding in educational scenes.