Inter-frame and Inter-layer Motion Prediction during Fast Block Motion Estimation in MCTF

  • Authors:

  • A. K. Karunakar;M. M. Manohara Pai

  • Affiliations:

  • Department of Information and Communication Technology, Manipal Institute of Technology, Manipal, India 576 104;Department of Information and Communication Technology, Manipal Institute of Technology, Manipal, India 576 104

  • Venue:
  • VISUAL '08 Proceedings of the 10th international conference on Visual Information Systems: Web-Based Visual Information Search and Management
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

The motion compensated temporal filtering (MCTF) using 5/3 LeGall bi orthogonal wavelet filter requires bidirectional motion estimation (ME) that is computationally intensive. Hence encoder of Scalable Video Coder(SVC) becomes very slow. This paper proposes Motion Prediction technique for Fast Block Based ME during MCTF in SVC. The technique exploits center biased characteristics of motion vectors. It applies coarse and fine search of any fast block based ME, only to the first pair of frames in a group of pictures (GOP). The generated motion vectors (MVs) are supplied to the next consecutive frames and even to subsequent temporal levels, that is used to initiate the fine search. The technique significantly reduces the number of blocks that undergoes ME in a GOP and hence reduces computational complexity of ME in MCTF. The proposed algorithm is implemented in MC-EZBC and there is no visible variations in the quality of the decoded video.