Realization of state-of-the-art intra-prediction in high efficiency video coding (HEVC)

Intra-picture prediction is an imperative part of High Efficiency Video Coding (HEVC/H.265) standard, which predicts the spatial correlation within a frame to remove redundancy. The intra-prediction process specified in HEVC/H.265 is computationally intensive and is relatively slow. Therefore, hardware implementation is desirable to accelerate the process. This thesis presents an efficient implementation of intra-prediction in an HEVC coder using MATLAB. MATLAB was chosen as it is easy to carry-over the implemented design to hardware by employing MATLAB’s HDL coder toolbox. Each of the 35 modes of intra-prediction are included in the design, viz. DC, planar, and 33 angular prediction modes. Intra-prediction is performed on the Luma as well as Chroma signal of an input image. Although implemented design works for any resolution of a video, the test images considered for the thesis are only of High-Definition (HD) or Ultra High-Definition (UHD) resolution. Quality analysis of the test benchmark is done using distinct criteria, namely visual quality, Structural Similarity Index (SSIM), and Peak Signal to Noise Ratio (PSNR). From the results obtained, it can be perceived that good spatial prediction is obtained for all images with excellent PSNR and similarity index. In addition, an analytical comparison of two state-of-the-art video compression standards, namely HEVC/H.265 and VP9, is also accomplished. The comparison is based on technical specifications and encoding performance of both standards.