Dust Removal with Boundary and Spatial Constraint for Videos Captured in Car

Abstract

Videos captured in car often suffer from duston the wind screen glass. The dust particles on wind screendecrease the video quality and make them blur. Removingdust and restoring high quality dust-free video is a challengingtask in the field of video stream processing. In this paper, we propose an improved iterative optimization pipeline toremove dust from the videos. Our method employs boundaryconstraint to keep transmission map in a reasonable rangeand use spatial constraint on the transmission map to avoidintroduction of significant halo artifacts into the resultantvideo. With optimized transmission map as an initial condition, our method can separate dust layer and background layer frominput video frames and keeps the background frames beingcolor faithful with fine details. Test results demonstrate thatour proposed method can recover dust-free videos from thedust contaminated input videos and keep the resultant video color faithful.