A FAST MOTION ESTIMATION ALGORITHM FOR DIGITAL IMAGE STABILIZATION

Abstract

The acquisition of digital video usually suffers from undesirable camera jitters due to unstable camera motion. The problem of stabilization of the video sequence from a camera mounted on a moving platform such as airborne reconnaissance, main battle tanks and ships is challenging and more complex. The motion composed of not only translation but rotation, zooming in and out and panning is a very common. Digital image stabilization is typically considered to contain mainly two main processes: motion estimation, and motion compensation. The motion estimation is the most computationally intensive part and thus plays an important role in the whole stabilization process. Many algorithms have been proposed in the literature to obtain fast or accurate motion vectors. Among the existing algorithms block Matching based full search is very common but it is highly computationally intensive. In this dissertation, we propose a fast motion estimation algorithm for digital image stabilization which reduces the computational cost of block matching based full search algorithm. Proposed algorithm not only reduces the complexity for estimation of translational motion but, rotational motion also. Simulations for the proposed digital image stabilization algorithm are performed on MATLAB. Some experimental results and comparisons with the existing methods for motion estimation is shown and discussed.