雾霾退化图像场景再现新算法

Novel method for haze degraded image scene rendition

为了快速有效地去雾,分析了单色大气散射模型和Retinex模型的对等关系,并从变分Retinex的角度提出了单色大气散射模型的求解新方法.求解过程首先构造了大气传递图的变分模型,将其估计问题转换为二次规划问题;其次,在3个颜色通道分别求解大气传递图,并反解单色大气散射模型得到反射图像;最后,依据颜色恒常性理论对去雾结果进行色彩复原.实验结果表明,新算法能够快速有效地去除浓雾.

To remove haze rapidly and effectively,equivalence between Retinex model and Monochrome Atmospheric Scattering model is explored,and then a novel solution for the Monochrome Atmospheric Scattering model is proposed based on the variational Retinex.Firstly,a variational model for the atmosphere transmission map is constructed,so the transmission map estimation problem can be formulated as a Quadratic Programming optimization problem.Secondly,we calculate the transmission map and resolve the Monochrome Atmospheric Scattering model in each channel to get the Reflectance image;Finally,the Color Constancy is adopted to restore the color of the dehazed image.Experimental results show that the proposed method could remove haze rapidly and effectively,and has some advantages over other algorithms.