Image stabilization by combining gray-scale projection and representative point matching algorithms

In this paper, a new digital image stabilization (DIS) algorithm base on gray scale projection algorithm (PA) and representative point matching (RPM) is proposed to stabilize videos. First, the gray-projection algorithm is used to estimate the global motion vectors (GMV). Second, we choose a block from the center of image, and estimate the local motion vectors (LMV) by use of RPM algorithm. Finally, we combine the two groups of vectors and calculate the final global motion vectors. The experimental results show that the proposed algorithm is better than traditional gray scale projection algorithm in accuracy, especially for the video sequences which contained interior moved objects. The algorithm can achieve good stabilizing accuracy with processing speed of 24 fps for 240×320 video sequences, it can meet the real-time processing requirement.