Meteorological approach for detection and removal of rain from videos

Rain removal algorithms require certain number of consecutive frames for reducing the visibility of rain from videos. These algorithms are designed for the rain videos captured by the fixed camera. In this study, it is demonstrated that using global motion compensation all the rain removal algorithms developed for fixed camera can be used for moving camera too. In this study a novel, efficient and simple algorithm for detection and removal of rain from video using meteorological properties is proposed. Here meteorological properties are used to separate the rain pixels from the non-rain pixels. The proposed algorithm achieves good accuracy in spite of less number of consecutive frames, which reduces the buffer size and delay. It works only on the intensity plane that further reduces the complexity and execution time significantly. Quantitative and qualitative analyses show that the proposed algorithm removes rain effectively under constrained buffer size and delay in comparison with most of the competing rain removal algorithms.