A Probabilistic Approach for Detection and Removal of Rain from Videos
Abhishek Kumar Tripathi, Sudipta Mukhopadhyay · IETE Journal of Research · 2011
AbstractA novel, efficient, simple, and probabilistic model based rain removal algorithm is proposed in this paper. Proposed algorithm is robust to the rain intensity variations. Probabilistic approach automatically adjust the threshold and effectively discriminate the rain pixels and non-rain moving object pixels. Discrimination between the rain and non-rain moving objects is based on the time evolution of pixels in consecutive frames. Proposed algorithm does not assume the shape, size and velocity of the raindrops and intensity of rain, which makes it robust to different rain conditions. For performance evaluation, in addition to miss and false detection a new metric temporal variance is introduced. Results show that proposed algorithm outperforms the other rain removal algorithms. Instead of working on all three color components, proposed algorithm works only on the intensity plane. Use of single plane reduces the complexity and execution time of the algorithm.