Spatial/temporal motion consistency based MERGE mode early decision for HEVC


Research output: Journal PublicationsJournal Article (refereed)peer-review

20 Citations (Scopus)


A framework for the early decision of MERGE mode (EDM) is proposed to overcome the brute force inter mode decision for the coding unit (CU) in High Efficiency Video Coding (HEVC). Firstly, the prediction-results of merge candidates (MrgCands) in previous and current frame are modeled to identify smooth/single-motion regions for EDM. Secondly, the RD-cost based statistical model is proposed for EDM since MERGE is selected 98.20% on average when 2N×2N and MERGE produce similar RD-cost. Finally, a model is proposed that obtains the SAD values for the prediction units (PU) of modes using the motion-estimation of 2N×2N and performs EDM based on the comparison of these SAD values. Experimental results demonstrate that an average 46.75% coding time can be saved, while the Bjøntegaard delta bit rate (BDBR) increase is only 1.79%. In addition, an average of 68.54% complexity reduction is achieved with 1.79% BDBR increase for conference video sequences.
Original languageEnglish
Pages (from-to)198-213
JournalJournal of Visual Communication and Image Representation
Publication statusPublished - 1 Apr 2017
Externally publishedYes

Bibliographical note

This work was supported in part by RGC General Research Fund (GRF) 9042322 (CityU 11200116); in part by the National Natural Science Foundation of China under Grants 61571274; in part by Shandong Natural Science Funds for Distinguished Young Scholar under Grant JQ201614; in part by the Young Scholars Program of Shandong University (YSPSDU) under Grant 2015WLJH39.


  • Fast MERGE mode prediction
  • Fast mode decision
  • High Efficiency Video Coding (HEVC)
  • Image coding
  • Inter mode decision
  • Merge candidates
  • SAD
  • Sum of absolute difference
  • Video coding


Dive into the research topics of 'Spatial/temporal motion consistency based MERGE mode early decision for HEVC'. Together they form a unique fingerprint.

Cite this