基于单尺度Retinex雾天降质图像增强算法

Fog-degraded Image Enhancement Based on Single-scale Retinex

摘要: 针对雾天采集的图像的退化现象，提出一种基于单尺度Retinex的雾天图像增强的算法。此算法首先把图像从RGB空间转换到 HSI空间，对亮度I分量用单尺度Retinex估计出与场景深度信息相关的雾气遮罩度，从亮度I分量中除去所估算的雾气遮罩度，即可得到亮度I分量的反射分量；然后利用双边滤波器，提取雾气遮罩度的反射分量，将该反射分量和亮度I的反射分量进行融合，并对饱和度S进行线性拉伸；最后转换到RGB空间即可得到增强后的图像。实验结果表明，该方法对雾天退化图像效果较好。

Abstract: According to degraded phenomena of images taken in foggy weather,this paper proposesd an algorithm of foggy image enhancement based on Single-Scale Retinex.The proposed algorithm first conversed RGB color space to HSI space.Using Retinex to estimate the fog veil of brightness whose distribution was in accord with the real fog density of the scene.The reflectance image could be obtained by subtracting the veil from the degraded image. Then,using bilateral filtering extracted the reflectance information included in the veil.The result image could be obtained by fusing the both reflectance images,using linear stretching to adjust saturation and conversing to RGB color space.Experiments show that the algorithm is more effective in foggy weather.