Adaptive exposure control and real-time image fusion for surveillance systems

This paper presents a surveillance system that can fuse two different exposed frames in real time and automatically track human faces as well as moving objects. Fusing multiple frames with different exposure time can accommodate the scenes with high dynamic range and the object tracking results guide auto exposure to ensure the right exposure for human faces and moving objects. The proposed flow as well as algorithms fully utilizes the calculation results getting from MPEG-4 compression task and the experimental result shows that the fusion speed is around 4 frames/seconds and the human face tracking speed is over than 10 frames/seconds