Low latency mitigation of rain induced noise in images

Abstract

The particles (rain drops and snow flakes) are usually large enough to be visible to camera. The particulates affect image Abstract intensity in a complex way and to some extent partially Sometimes, weather conditions, paocclude the necessary image information (see Figure 1).Sometimes, weather conditions, particularly rami and snow, Studies in the field of atmospheric sciences, especially remote makes it difficult to obtain clear image or videos Partial sensing, show that the affect of particles in dynamic weather occlusion by moving rain drops makes it difficult to process are predictable [2,12,9]. such videos for example in object recognition and feature Dynamic weather affects have been removed to some extent tracking. The difficulty further increases in presence of by use of temporal filtering [11]. However, this causes camera and objects motion. Motion analysis is typically bluffing especially when camera or objects move. Zhang et al performed to avoid distortions in camera and object motion. [13] suggest correction for camera motion prior to rain This paper proposes an image processmig procedure to carry removal. The idea is to extract camera motion, by frame out rain removal in image sequences without explicit motion alignment, before processing and once processed for rain analysis. The proposed method analysis temporal changes removal the extracted motion is injected again in the video. (with l frame latency) and selects rain corrupted image Frame alignment may not be possible for example in presence pixels. The selected pixels are further filtered for camera and of object motion. Extraction of motion information in object motion by using spatio-temporal consistency presence of rain is difficult [1] and may produce unreliable constrains. The corrupted pixels are recovered by intensity results. In more recent research, Garg and Nayar [6] suggest subtraction. A naive selection of algorithms shows segmentation of individual rain streaks. They assume comparable results. The proposed scheme has various stationary background within a predefined spatial locality. applications for example outdoor surveillance, feature This allows removal for limited dynamic scenes, however, tracking and object recognition. fails when rain is seen against a moving textured background.

Reserch on Removal Technology of Snowflakes in Video Images

“…Barnum [4] et al introduced the global characteristics of rain and snow, and proposed the removal method using frequency-domain characteristics of rain and snow. Subhani [5] et al detected rain by area frame gray difference, but the drawback is that it can not achieve the desired effect when the moving object texture is more complicated or the raindrops is too large.…”