Motion filter for digital image stabilization

Desen Yin,Yuejin Zhao,Bin Wang
2008-01-01
Abstract:A novel algorithm based on particle filtering is proposed for digital image stabilization. Local motion vectors are estimated by block matching, and the global motion parameters computed by random least square method. Meanwhile, particle filtering is employed to facilitate smooth operation in digital image stabilization. Motion compensation is completed. Experimental results for many video sequences showed that the method can eliminate high frequency trembles, sustain active motion of camera and realize real time image stabilization satisfactorily.
What problem does this paper attempt to address?