改进的异质大气光估计的图像去雾算法

摘要

为解决在雾霾天气中捕获的图像不清晰的问题,提出一种改进的异质大气光估计法及一种基于非线性颜色衰减先验模型精确估计场景深度法,进行单一图像去雾处理.利用均值池化的思想对HSV颜色空间的亮度成分进行处理,估计异质大气光,该方法不依赖于某一具体像素值,估计出的异质大气光鲁棒性更好;提出一种非线性颜色衰减先验模型,能克服以往线性颜色衰减先验模型估计场景深度时出现负值的问题.实验结果表明,所提算法达到甚至超过了当前先进的图像去雾方法.%To solve the problem that the images captured in the haze weather are not clear,an improved heterogeneous atmosphere light estimation method and a depth estimation algorithm with color attenuation prior (CAP) were proposed for dehazing single image.The brightness component of the HSV color space was processed through the idea of mean-pooling,and the heterogeneous atmosphere light was estimated,which was more robust because of its independence of a specified pixel.Nonlinear CAP model was proposed for estimating the scene depth,which overcame the defects of the occurrence of negative scene depths from the linear CAP model.Experimental results show that the proposed algorithm outperforms the current advanced methods in dehazing images.