Digital image stabilization based on statistical selection of feasible regions

This paper presents a novel and robust approach for digital image stabilization system based on a statistical region analysis for dominant motion selection. The proposed stabilization system adaptively employs the feature-histogram for feasible region selection resulting in stabilization of undesired camera fluctuation regardless of object's movement. The main difference between the previous and proposed approaches is that existing state-of-the-art methods use points, edges, or bit-plane patterns to select the feasible regions, whereas the proposed algorithm employs a feature-histogram. Experimental results show that the proposed algorithm improves the stabilization performance by 7% on average with four times less computation than the existing methods by skipping motion estimation and feature detection in disqualified blocks. Therefore, the proposed algorithm is suitable for consumer applications such as video camcorders and digital cameras.