H.264/AVC-to-SVC Temporal Transcoding using Machine Learning

Nowadays, networks and terminals with diverse characteristics of bandwidth and capabilities coexist. To ensure a good quality of experience, this diverse environment demands adaptability of the video stream. In general, video contents are compressed to save storage capacity and to reduce the bandwidth required for its transmission. Therefore, if these compressed video streams were compressed using scalable video coding schemes, they would be able to adapt to those heterogeneous networks and a wide range of terminals. Since the majority of the multimedia contents are compressed using H.264/AVC, they cannot benefit from that scalability. This paper proposes a technique to convert an H.264/AVC bitstream without scalability to a scalable bitstream with temporal scalability in Main Profile by accelerating the mode decision task of the SVC encoding stage using Machine Learning tools. The results show that when our technique is applied, the complexity is reduced by 87% while maintaining coding efficiency.