Motion-estimation-based stabilization of infrared video

Abstract

In the course of the filming of infrared (IR) video, intrinsic equipment instability incurs movement that in turn causes image blurring. For image clarity and viewing comfortability, it is required that such movement be countered. Presently, video stabilization systems perform Motion Estimation of frames that is then applied frame-by-frame to subsequent frames in order to calculate a motion vector, counter movement, and produce, thereby, a more stable image. However, frame-by-frame comparison for long-distance filming often is difficult due to lack of information. The present study determined the appropriate blocks with the most information for Motion Estimation. We also were able to differentiate between equipment movement and movement in the video itself. By these means, we were able to stabilize videos. The methods employed in the experimentation were 5 sets of 640 × 480 long-distance videos and 5 sets of 480 × 320 long-distance videos. When compared with the current motion estimation methods, our proposed method afforded a 10% increase in accuracy.