Affiliations: [a] Department of Computer Science, University of Kerala, Kariavattom, Thiruvananthapuram, Kerala, India
| [b] Department of Electrical Engineering, University of Texas at Arlington, TX, USA
| [c] Department of Computer Science, Cochin University of Science and Technology, Kochi, Kerala, India
Correspondence:
[*]
Corresponding author. Preethi S. Nair, Department of Computer Science, University of Kerala, Kariavattom, Thiruvananthapuram-695581, Kerala, India. E-mail: preethi.snair85@gmail.com.
Abstract: The state-of-the-art video compression standard, High Efficiency Video Coding (HEVC) has standardised to contemplate a better compression of high and ultra high resolution videos. HEVC introduced a lot of new coding tools to meet its improved coding efficiency, but at a bit increase in the computational cost. The intra prediction mode decision strategy is one among them and the use of a large number of intra prediction modes at different PU sizes is the reason for an improved mode prediction as well as the high computational complexity. Besides the number of prediction modes, HEVC adopts a Rough Mode Decision (RMD) process as well as Rate Distortion Optimization (RDO) process in the intra prediction stage, which takes substantial amount of execution time. Hence a method to reduce this time complexity by incorporating machine learning technique in RMD process is proposed in this work. In this method, we use Support Vector Machine (SVM) to find out the best mode in the RMD stage. Experimental results indicate that the proposed method considerably reduces the computational cost of the HEVC reference software while retaining the visual quality of videos. Since HEVC supports real-time video processing, the proposed method sounds to be applicable.
