Abstract
Screen content coding (SCC) is the extension to high-efficiency video coding (HEVC) for compressing screen content videos. New coding tools, intrablock copy (IBC), and palette (PLT) modes, are introduced to encode screen content (SC) such as texts and graphics. The IBC mode is used for encoding repeating patterns by performing block matching within the same frame, while the PLT mode is designed for SC with few distinct colors by coding the major colors and their corresponding locations using an index map. However, the use of IBC and PLT modes increases the encoder complexity remarkably though coding efficiency can be improved. Therefore, we propose to have a mode skipping approach to reduce the encoder complexity of SCC by making use of SC characteristics, neighbor coding unit (CU) correlations, and intermediate cost information via random forest (RF). Detailed feature analyses and sample preparation are also described. A novel hyperparameter tuning approach with the consideration of coding bitrate and encoding time is proposed for RFs at each CU size to further boost the encoding process. Experimental results show that our proposed approach can obtain 45.06% average encoding time reduction with only a 1.08% increase in Bjontegaard delta bitrate. Average encoding time can even be reduced to 58.57% by regulating the hyperparameters.
Original language | English |
---|---|
Article number | 8673881 |
Pages (from-to) | 2433-2446 |
Number of pages | 14 |
Journal | IEEE Transactions on Multimedia |
Volume | 21 |
Issue number | 10 |
DOIs | |
Publication status | Published - Oct 2019 |
Keywords
- HEVC
- machine learning
- random forest
- screen content coding
- video coding
ASJC Scopus subject areas
- Signal Processing
- Media Technology
- Computer Science Applications
- Electrical and Electronic Engineering