Recovery Of Video Degraded By Rain Based On Temporal And Spatial Characteristics

Outdoor weather factors make a lot of informations of video be covered or blurred inrainy days. Many detail informations were not recognized.Those affect the normal workand the efficiency of the video system,and even cause the system paralyzed. Eliminatingthe influence of negative factors to improve the quality of visual video have becomeparticularly important.Rainy days is a complex weather, the spatial and temporal distribution of high-speedfalling raindrops is random, which makes the rain scene model become complicated, andspecific physical model is only adapt for the specific rain scene.This paper uses a partial fenestration on the brightness to detect rain area. Thismethod determine the difference of continuous frames, then open a proper window for thedifference image. This algorithm can distinguish scenes and moving target effectively anddetect rain area which within the moving target area. In additionally, the algorithm doesnot do a complicating registration process for moving targets, which improving the speedof the system. Adding characteristics of rain line such as size, orientation restrict the initialrain, then completing the detection of the rain zone. The detection process uses thecharacteristics on brightness, size, orientation, it can get the precise rain zone.After the Detection, an approach base on the curvature-driven diffusion model is usedto the removal of the rain. This is an improve model basing on the total differential.Curvature-driven diffusion model can achieve a good visual effects for single frame image,but it flicker in the entire video. A smoothing processing using median filter for framescan eliminate flicker. Detection of rain based on the spatial and temporal characteristics ofthe rain line, it is effective for the detection of small rain scenes, but it is inadaptable forheavy rain scenes.