Robust Salient Moving Object Detection with Light-Computational Load

In this paper, we propose a method which detects salient moving objects with light-computational load. Though Gaussian Mixture Model is widely used for object detection, it is computationally heavy. On the other hand, basic methods like temporal difference are simple and fast but they have constraints as hole or ghost problems. We have combined these algorithms to overcome each one's weakness. We use background modeling and subtraction method which are similar to adaptive threshold with foreground map. Foreground map is generated by Modified Temporal Difference to speed up the process. Using adaptive threshold, we have improved the performance, when there is slightly moving background like branches in the wind. So we can eliminate meaningless objects with lightcomputational load. Experimental results show efficiency and robustness of our algorithm in several outdoor scenes.