Research On Restoration Of Video Under Rain Condition

In rainy days, images of outdoor scenes received by the camera are usually degraded bythe weather. Many features in the image are covered or blurred, and the details of theimage cannot be recognized. This greatly limits the effectiveness of outdoor vision systems,in some cases may even cause more serious consequences. Therefore, the research on how toeliminate the impact of rain of outdoor video to improve the visibility of the scene has avery important use value.Nowadays, in the research of restoration of video under rain condition, the generalresearch model is like this: Firstly, it’s necessary to analyze the performance of rain in theimage preciously, and find many special characteristics of rain; Secondly, a model about thecharacteristics of rain is established to detect rain; Finally, the detected rain is removed byreplace it with the pixels in the two neighboring frames (on either side) that are not effectedby raindrops. However, these methods are only applicable to static scenes or the situationthat moving objects in the scene is in the room (that is to say the moving objects were notaffected by rain). In practical applications (such as Video Surveillance), movingobjects outdoor will be affected by rain definitely. And it’s natural to pay more attentionto the detection and restoration of rain which affect moving targets.Focusing on the restoration of moving targets degraded by rain in the image, a detectionapproach based on local brightness characteristic of rain was proposed in this paper. Firstly,to judge each pixel whether was affected by rain, a local window centred at this pixel in thebrightness subtraction image between the adjacent video frames was established and theproperty of the pixel was judged based on the brightness characteristic of rain. This methodcan detect all the rain in the scene at a time.Secondly, a new way to describe the direction ofrain was adopted. And based on the brightness preliminary detection, the constraint ofdirection of rain was added to remove the false alarm and got the accurate location of rain.The final restorated image with enhanced contrast was obtained by performing frame repairbased on space structure information. And in order to eliminate the flicker in the video, amedian filtering between frames was applied. For the case of heavy rain, the video imagewas seriously affected by fog. Focusing on the phenomenon, a foggy image restorationMethod was added to the removal rain based on Frequency characteristic of rain algorithm,which greatly increased the contrast and visibility of the scene.