Algorithm research on motion area segmentation based on improved frame differences

Abstract

Frame difference is most widely used in video detection, but the original frame difference is not accurate enough to segment the moving object in special scenes such as light changes suddenly or camera shakes slightly. In this paper, we propose a segmentation algorithm based on improved frame differences. In the algorithm, we first set the frame difference value k by the environment characteristics, and then we determine whether the image movement is a global movement by the dual-threshold. If the global movement is caused by the illumination variation, we calculate the changing rate of gray value to detect the moving object. If it is caused by the camera shaking, we find the main direction of the movement by comparing a point in the background with the eight neighborhood points in the current frame to detect the moving object. The simulation results show the effectiveness of the proposed algorithm.