A robust video stabilization using optimized dynamic time warping

Video stabilization has become a demanding field with the emerging compact and handy imaging devices in the market. Videos recorded with these devices mostly suffer from some unwanted jittery motions induced due to shaky hands or environmental vibrations. This paper presents an efficient Dynamic Time Warping (DTW) approach for the translational jitter stabilization in the recorded videos. The approach utilizes the dynamic programming for the optimized performance. The use of dynamic programming based DTW gives a significant improvement in the processing speed and memory consumption over the classical DTW approach. The algorithm has been tested on different categories of video like moving platform, homogeneous regions and low quality blurred videos, which are generally considered as bottleneck problems for accurate motion estimation. The proposed approach provides better stabilization in relation to the existing intensity based methods and the computational efficiency is inherited from recursive property of dynamic programming. Comparative analysis of results is given in terms of difference frame analysis and overall performance evaluation is given using interframe transformation fidelity (ITF) factor and processing time.