Research On General Fog-removal Framework And Algorithm

The performance of outdoor monitoring imaging system is greatly limited byfactors such as weather, light. In the fog weather, the image has the seriousdegradation phenomenon due to the scattering effect of the particles in theatmosphere. It will cause image blur, faint colors, poor contrast and detailsinformation loss, which will be serious influence on the applications value of theimage. In order to reduce the impact of fog on the imaging system, we firstly analyzethe physical model of fog imaging and the reason of image degradation, summarizeand discuss the basic theories and kernel technologies of classical fog-removalalgorithm. Research and implementation in this paper are as follows:（1） This paper has analyzed the deficiency of He’s algorithm, and thenproposed the improved algorithm: use of the maximum filter instead ofthe repair module in He’s algorithm, which took the greatest of processingtime. The experiments show that the new algorithm can increase thespeed by10to20times under the effect of ensuring fog-removal.（2） The Depth-Consistency Principle and The Space-Continuity Principle(The DCP and SCP) are proposed through analyzing the physicalcharacteristics of the fog distribution. Above two principles, we proposedthe general fog-removal framework, and applied a variety of filtercombination to the framework. And we achieved a good recover results,which proves the validity of two principles and the generality of theframework.（3） By simplifying the bilateral filtering, we propose the fast fog removalalgorithm base on the DCP and SCP. And study the optimization methodof the algorithm: we used integer lookup table for Range filtering, andused the optimization method of integral figure in the average filteringalgorithm, further improve the processing speed and real-time of thealgorithm through CNC comprehensive assessment system.（4） In view of the recovered image after fog removal has the problem of low overall brightness, this paper presents a simple and effectivepost-processing module: adopt the method of linear strengthen to improvethe values of each pixel in the image, and it can improve the image visualeffect, and keep the image without color deviation.This research project finally have the fog algorithm into the dynamic library file,use array pointer operations to improve the execution speed of algorithm performance,and finally program an images and video demo to realize fog removal.