Speed up HEVC Encoding Process by Cloud-Computing and Early Termination Method

Autor: Sheng-Yan Wang, 王聖諺
Rok vydání: 2014
Druh dokumentu: 學位論文 ; thesis
Popis: 103
With the advance of video coding technology, the resolution and quality of video frames have been largely improved. However, high resolution and high coding quality also imply high coding bitrate and higher time complexity. The high efficient video coding standard (HEVC), approved by ITU-T, in April 13, 2013, provide twice the compression efficiency, as compared to its previous version, H.264/AVC. The processing resolution range can be from 4K ultra HD up to 8192x4320. The HEVC was also designed to enable parallel processing to provide widespread coding platform applications, such as Wavefront Parallel Processing (WPP), independent tile coding, and entropy slice coding etc. In addition, the HEVC supports different processing units, such as Coding Unit (CU), prediction unit (PU) and transform unit (TU). In comparison with H.264/AVC, whose coding unit is fixed as 16 by 16, the HEVC supports CU from 64x64 to 8x8. The CU can be split asymmetrically to reduce encoding bit rate. However, these high efficient coding procedures also demand higher computation complexity. In this these, we proposed to improve the HEVC coding speed by utilizing cloud computing platform and multi-thread processing, in additional to developing early termination methods: (1) we segment the video into GOP units with different length to distribute to different cloud computing nodes; (2) The video coding can be speedup by parallel processing by dividing the image frame into tiles, such that multi-thread computing can encode them independently; (3) By referencing to surrounding CUs and Pus, the current CU depth decomposition and PU prediction can be early terminated to eliminate extra computational operations. In total, the proposed HEVC speedup coding methods can save 94.813% encoding time, while slightly increasing the bitrates and decreasing the PSNRs, as compared with the HM13.0 HEVC encoder description.
Databáze: Networked Digital Library of Theses & Dissertations