నైరూప్య

A Rain Pixel Recovery Algorithm for Videos

Bhuvaneshwari C.Melinamath Amruta S. Bandarwad

Rain removal is a highly useful and important technique in many applications such as surveillance and movie editing. In the recent years several rain removal algorithms have been proposed, where photometric, chromatic, and probabilistic properties of the rain pixels have been utilized to detect and remove the rainy effect. The existing methods generally work well with light rain and relatively static scenes, but when dealing with heavier rainfall in dynamic scenes in motion, these existing methods give very poor results. The proposed algorithm is based on the concept of motion segmentation of dynamic scene. After applying photometric and chromatic constraints for rain detection, several rain removal filters are applied on the rainy image such that their dynamic property as well as motion occlusion clues are taken into consideration; both spatial and temporal information are then adaptively exploited during rain pixel recovery. Extensive simulation results show that the proposed algorithm shows a much better performance for rainy scenes with large motion than the existing algorithms.