Detection and compensation of image sequence jitter due to an unstable CCD camera for video tracking of a moving target

Abstract

Nowadays image motion analysis is considered as a significant subject in image processing and according to its characteristics is divided into different categories. This paper focuses on camera jitter as fast translations in image sequence due to an unstable platform. Our studies have led to do correlation matching [C-S. Fuh et al. (1991)] between well featured templates [M.A.Z. Pisheh et al. (2003), Y-P. Tan et al. (1996)] of successive frames to prevent aperture problems and thus have good estimation of the translations. But camera vibration may also suffer from rotation and scaling as nonlinear motions [G.J. Foster et al. (2000)] that may defect the matching process. Therefore mapping of 3D image points into camera image plane has been investigated to trace the effects of these nonlinear parameters in 2D motion equations [F.V. Heijden, (1996)]. Now by approximating these equations for minute rotation and bounded scaling between frame i and i+1, the set of motion equations with four unknowns ( rotation, scaling, horizontal/vertical translations) could be solved just by LSE [M.S. Grewal (2001)] method for a unique and optimized response. After all when we apply the estimated displacements on each frame for compensation, an stabilized sequence will conclude.