Evaluation of video stabilization metrics for the assessment of camera vibrations

For various applications, such as handheld imaging systems and cameras mounted on vehicles or flying platforms, unavoidable motion and vibrations of the camera may result in smeared or blurred images or shaky videos. This study aims at evaluating metrics for the measurement of camera vibrations in image sequences considering triangles and bars as test patterns. The focus are objective metrics for video stabilization, which are designed to objectively evaluate whether video stabilization was able to eliminate objectionable visual movement. The metrics are evaluated for simulated image sequences captured by an artificially moved camera. The sequences vary in different properties such as the sensor noise of the camera, as well as the and temporal frequency of the camera vibrations. We analyze the effect of these properties on the metrics behavior. First results using recorded data of thermal imagers are presented as well. The findings will provide insights into the efficacy of different video stabilization metrics on simulated sequences with varying properties.