영상 감시를 위한 움직이는 물체의 그림자 제거 기법

Moving object detection, which separates moving objects from video sequences, is essential to video surveillance However, the separated moving objects often contain shadows with it. These shadows can confuse to localize the exact position of the object and also they makes it difficult to distinguish different objects. In this paper, a method to eliminate shadows from foreground image is proposed. The proposed method models the probability that a pixel is belonged to the shadow area based on the characteristics of shadows. The threshold level for determining shadow pixels is set adaptively to cope with the change of illuminations. In the experiments, the proposed method effectively eliminated shadows and it was able to process 30 frames per second.