An Image Stabilization Algorithm for Intra Frame and Inter Frame of Video

Since the airship-borne camera recorder is easily effected by the complex weather conditions such as fog and haze etc, we adopt the multiple scales Retinex algorithm with color restoration to remove fog. It retains the details of the image enhanced and restores the genuine color of the image. And then the algorithm based on SIFT feature extraction and match in the Euclidean space is utilized to estimate the global motion and vibration component. In order to make sure the precision of global motion estimated, we adopt the RANSAC technique to delete the front motion objects match. The adaptive median filter is proposed to eliminate the high frequency vibration component, and the filter window could be adjusted according to the amplitude of vibration component. Finally, the image mosaic technique is utilized to reconstruct the undefined areas which appears in the process of global motion compensation. The experiments show that the algorithm proposed has very robust removing fog and anti-vibration abilities, it achieves the video stabilized smooth, continuous and seamless output.