Image Dehazing Study Based On Structure Preserving

Restoring the actual scene is a challenging problem in computer vision. In fogweather conditions, due to the effect of scattering of the particles in the atmosphere,That will be serious mass reduction by making access to the image to a certain extent.Not only the visibility of the blurred image, but also for the extraction of imagefeatures, image real information has a great disturbance. Haze weather often affectsthe normal operation of the monitoring system of outdoor, produces relevant aspectsof the people engaged in the work of safety concerns. Therefore, It has very importantpractical significance how to use effective and quick way to restore atomizationimage.This paper proposes a novel dehazing method from single image based onstructure preserving. Different from most existing methods, the structure informationis considered sufficiently for visibility enhancement. First, the initial airlight isgenerated by filtering the minimum channel of haze image, meanwhile the structureinformation of the minimum channel is extracted as the reference image. Then, theinitial airlight is refined with the reference image by guided joint bilateral filter, whichguarantees the scene radiance and depth more naturally. Finally, the actual image isrestored by solving the atmospheric attenuation model. Specifically, an improvedmethod based on quad-tree subdivision is presented to obtain an accurate atmosphericlight. We verify the effectiveness and feasibility of our method on series of real hazeimages. Experimental results indicate that the proposed method can achieve gooddehazed results at a distant scene. In particular, compared with several well-knownmethods, our method exhibits much better performance for single image dehazing.