Improved Video Stabilization using SIFT-Log Polar Technique for Unmanned Aerial Vehicles

The video taken from unmanned aerial vehicles are usually not stabilized by dynamic conditions of the camera platform. This paper presents an improved solution for video stabilization of aerial vehicles by combining "scale invariant feature transform (SIFT)" with "log-polar transform". The input video frame is transformed into log-polar that provides extended scale and rotation invariance. Afterward, key-points are extracted using the SIFT method and tracked in consecutive frames to eliminate the unwanted motion in a video. Hence, the presented technique performs better even if the object's scale and orientation vary in video frames when unmanned aerial vehicles change its altitude and orientation during the flight. The results of SIFT and log-polar method are compared with the standard SIFT-ME algorithm to analyze the scale and rotation invariance capability for video stabilization. The proposed technique shows the better capacity to stabilize video even when the UAV changes its attitude and orientation (i.e. scale change tolerance: 44.5% and Rotation tolerance: 53%).