Fast motion estimation for real-time shape-adaptive MPEG-4 encoding

Authors:
Peter Kauff;Klaas Schüür
Affiliations:
Heinrich-Hertz-Institute (HHI), Image Processing Department, D-10587, Berlin, Germany;Heinrich-Hertz-Institute (HHI), Image Processing Department, D-10587, Berlin, Germany
Venue:
MULTIMEDIA '00 Proceedings of the 2000 ACM workshops on Multimedia
Year:
2000

Citing 2
Cited 0

Performance of optical flow techniques

International Journal of Computer Vision
Variable-Raster Multiresolution Video Processing with Motion Compensation Techniques

ICIP '97 Proceedings of the 1997 International Conference on Image Processing (ICIP '97) 3-Volume Set-Volume 1 - Volume 1

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents a fast motion estimator which can be used for real-time MPEG-4 encoding of arbitrarily shaped video objects. The approach is based on an existing algorithm which has already been applied successfully to format conversion. To exploit it for shape-adaptive coding, the algorithm has been adapted to the special properties of the MPEG-4 standard. With this new tool it becomes possible to encode arbitrarily shaped video objects (CIF, 25 Hz) in real-time with a MPEG-4 software encoder at a Pentium III 500 MHz. The real-time capability has to be paid by a slight loss of coding efficiency (about 0.2 dB in terms of rate-distortion measurements), compared to the MPEG-4 verification model.