Collocated macroblock based motion estimation for H.264/AVC on GPU

Research output: Journal Publications and ReviewsRGC 22 - Publication in policy or professional journal

2 Scopus Citations
View graph of relations

Author(s)

Related Research Unit(s)

Detail(s)

Original languageEnglish
Pages (from-to)15-18
Journal / PublicationHKIE Transactions Hong Kong Institution of Engineers
Volume17
Issue number3
Publication statusPublished - Sept 2010

Abstract

Recent results show that the parallel processing capability of the powerful Graphics Process Unit (GPU) can greatly accelerate motion estimation in H.264/AVC encoder implementation. However, the bit rate cost of motion vector (MV) is always ignored in this parallel motion estimation implementation on GPU due to the unavailability of the spatially predicted MVs, which accordingly degrades the rate-distortion performance. In this letter, a novel collocated macroblock based motion estimation algorithm is proposed to tackle this problem by using temporally predicted MV. The experimental results demonstrate that the peak signal-to-noise ratio (PSNR) improvement can be up to 0.8 dB with a negligible computational complexity increase.

Research Area(s)

  • Compute unified device architecture (cuda), Graphics processing unit (gpu), H.264/AVC, Motion estimation, Parallel processing