Motion Stabilization of Aerial Infra Red Videos

Abstract

DIGITAL image sequences captured from cameras attached with aerial vehicles such as helicopters display unwanted high frequency jittering motion. The amount of motion present in an image sequence depends on the physics of the motion capture device relative to objects in the captured sequence. Such applications usually require an image stabilization system aiming to produce a compensated video sequence so that the image motion due to the camera’s undesirable shake or jitter can be removed. The aim of our project is to build low cost software solution to stabilize such videos. The project contains four major modules i.e. the feature tracker module, homography computer, mosaic creator and stabilized frame generator. The feature extractor module uses SIFT algorithm for the extraction of feature points from the frames of the video. The feature points of one frame is matched with the feature points of another frame or previously created mosaic. These matched feature points are then used for the computation of homography matrix which represents the transformation required for one frame to be aligned with the other frame. The homography matrix is then used for the creation of mosaic. Reference positions of the frames with respect to the first frame are saved and are used to make the stabilized frame for the new stabilized video. The stabilized video can then be used for surveillance purposes in military and in making different movies or as an input in many different applications.