基于时空特征统计学习的视频图像雨滴去除问题研究

In this paper, the method of rain removal from video images is proposed for video surveillance system. Basing on the physical properties and time domain characteristics, it uses the Kalman Filter Algorithm to remove the raindrops directly from video images in the static scene by tracking the image background. While in the dynamic scene, it presents an algorithm based on the statistical learning of spatial-temporal property, which accurately distinguish the rain and moving pixels, exclude the interference of moving objects and improve the quality of video images effectively. Experiments show that the proposed algorithm produce effective results and calculate fast which can meet the real-time requirements of system.