Using Edge Detection to Accelerate HEVC Mode Decision on GPU

碩士 === 國立清華大學 === 資訊工程學系 === 104 === HEVC is the latest video compression standard, which provides better compression ration and video quality than H.264. However, its computational time is much higher than previous standards. One of the performance bottleneck is mode decision, which searches the be...

Full description

Bibliographic Details
Main Authors: Wang, I Ching, 王儀靜
Other Authors: Lee, Che Rung
Format: Others
Language:en_US
Published: 2016
Online Access:http://ndltd.ncl.edu.tw/handle/69293212297629205389
Description
Summary:碩士 === 國立清華大學 === 資訊工程學系 === 104 === HEVC is the latest video compression standard, which provides better compression ration and video quality than H.264. However, its computational time is much higher than previous standards. One of the performance bottleneck is mode decision, which searches the best mode for motion estimation. In this thesis, we proposed a method to accelerate the mode decision and motion estimation on the CTU level. The method uses edge detection to estimate the proper mode of each CTU blocks, instead of searching the best mode from bottom to top. The method has been implemented on GPU and CPU. Experiments show that comparing with the x265 and previous work, the performance can be improved up to eight times with slightly increases in bitrate and PSNR.