VIDEO TARGET STABILIZATION BASED ON MOTION VECTOR COMPENSATION

Abstract

In this paper a solution is proposed for stabilizing the target in un-stabilized shaky videos by using the motion vector compensation technique mainly focusing on template-based approach for locating target in desired frames. The input video frames are taken and are converted to the appropriate required format. Necessary and appropriate noise filtration is usually done to neutralize the affects of different noises on the video. Necessary basic information and parameters are extracted from the input video / frames. Target is manually selected and matched in successive frames by using cross correlation matching. Center positions are then updated which are later used for motion vector calculation and offset updating.  Finally, the effects of un-necessary target and camera movements are removed by necessary linear translations. Stabilize frames are then aligned in sequence for video reconstruction and display. The proposed algorithm is useful in security applications and in research works like astrological studies and the study of heavenly bodies.