【作者】 叶炯凯； 武勇； 彭冬亮；

【Author】 Jiongkai Ye;Yong Wu;Dongliang Peng;The School of Automation,Hangzhou Dianzi University;Zhejiang Institute of Communications;

【机构】 杭州电子科技大学自动化学院； 浙江交通职业技术学院；

【摘要】 目前的图像去雾算法针对图像阴影部分往往存在对比度下降和颜色失真的问题,如何有针对性的对图像有雾和暗光部分进行增强是解决问题的关键。研究者基于大气模型提出了一个同时考虑有雾和暗光的完全大气模型,该模型在大气散射模型的基础上将黑暗(全球黑暗因子)作为一种干扰因素,通过估计全球大气光和全球黑暗因子实现了自适应的亮度平衡。研究者根据物体显色原理解释了暗通道先验去雾的合理性,延伸出亮通道先验微光增强算法。实验结果表明该方法能够很好的改善图像对比度下降和颜色失真问题,说明多任务同时去雾和微光增强的视觉效果要优于单一任务的去雾。更多还原

【Abstract】 The current image dehazing algorithms often suffer from contrast reduction and color distortion in the shadow regions of the images.Targeted enhancement of hazy and low-light regions in images is crucial for addressing the problem.The researchers proposes a complete atmospheric model that considers both haze and dark light simultaneously,which includes darkness(global darkness factor)as a disturbance factor on top of the atmospheric scattering model as well,and achieves an adaptive luminance balance without the need to estimate the global atmospheric light and the global darkness factor.The researchers explained the rationality of the dark channel prior dehazing based on the principle of object color rendition,extend the bright channel prior for low-light enhancement.The experimental results demonstrate that this method effectively improves the issues of decreased image contrast and color distortion,indicating that the visual effects of simultaneously removing haze and enhancing low-light are superior to single-task haze removal.更多还原