Implementation of fast residual quadtree coding and fast intra prediction in high efficiency video coding.

摘要

High Efficiency Video Coding, one of the latest video compression schemes proposed by Joint Collaborative Team on Video Coding (JCT-VC) which is a partnership between International Telecommunication Union Video Coding Experts Group (ITU T-VCEG) and International Organization for Standardization Moving Picture Experts Group (ISO/IEC MPEG), has been promoted to Final Draft International Standard (FDIS) on January 25, 2013. With three profiles: main, main10 and still-frame, it achieves far better compression ratio than already existing schemes, while maintaining visual quality. However, this superior bitrate savings is achieved at the expense of huge computational complexity which in turn means higher encoding time. One of the aspects of encoding is intra prediction. A fast intra prediction method has been explored here. Residual quadtree (RQT) coding is another computation intensive step before intra prediction. Here, a fast RQT method has been combined with the fast intra prediction method to achieve better encoding time. In this thesis, a decrease of about 50% in encoding time can be seen with combining the fast intra prediction method and fast RQT into a single codebase for encoding the raw YUV videos while maintaining the visual quality.