High-Level Synthesis Based VLSI Architectures for Video Coding

Abstract

High Efficiency Video Coding (HEVC) is state-of-the-art video coding standard. Emerging applications like free-viewpoint video, 360degree video, augmented reality, 3D movies etc. require standardized extensions of HEVC. The standardized extensions of HEVC include HEVC Scalable Video Coding (SHVC), HEVC Multiview Video Coding (MV-HEVC), MV-HEVC+ Depth (3D-HEVC) and HEVC Screen Content Coding. 3D-HEVC is used for applications like view synthesis generation, free-viewpoint video. Coding and transmission of depth maps in 3D-HEVC is used for the virtual view synthesis by the algorithms like Depth Image Based Rendering (DIBR). As first step, we performed the profiling of the 3D-HEVC standard. Computational intensive parts of the standard are identified for the efficient hardware implementation. One of the computational intensive part of the 3D-HEVC, HEVC and H.264/AVC is the Interpolation Filtering used for Fractional Motion Estimation (FME). The hardware implementation of the interpolation filtering is carried out using High-Level Synthesis (HLS) tools. Xilinx Vivado Design Suite is used for the HLS implementation of the interpolation filters of HEVC and H.264/AVC. The complexity of the digital systems is greatly increased. High-Level Synthesis is the methodology which offers great benefits such as late architectural or functional changes without time consuming in rewriting of RTL-code, algorithms can be tested and evaluated early in the design cycle and development of accurate models against which the final hardware can be verified.