Global Motion Estimation for Extremely Low Resolution Video Images

E. Nakamura, I. Sugihara, and K. Sawada (Japan)


Real time video image processing, low resolution video, motion estimation, elementary motion detectors, motion parallax


This paper presents a global motion estimation algorithm for extremely low resolution video image sequences. The primary contribution of the algorithm is to estimate direc tions of global motion vectors even when extremely low resolution images of size such as 18 12 are input to the algorithm. The estimation of global motion vectors are per formed based on the calculation of local motion vectors, which are computed by hiring so called elementary mo tion detectors or EMDs. Our EMD based algorithm, im plemented in a software package using Microsoft's Direct Show modules running on a PC platform, has found to pro vide far better performance than a block matching method in terms of direction estimation accuracy when extremely low resolution vide images are considered.

Important Links:

Go Back