一种基于单尺度Retinex的雾天降质图像增强新算法

Fog-degraded image enhancement based on single-scale Retinex

雾图增强是智能管理系统的一个组成部分,在交通管理系统、公路收费站、轮船、飞机场等场合有着广泛的应用场景。比较了对数函数、双曲正切函数、反双曲正切函数对雾天降质图像亮度分量的调节能力,证明双曲正切函数具有比对数函数更宽的亮度调节能力。在此基础上提出一种基于单尺度Retinex的雾天图像增强方法。该算法首先把图像从RGB彩色空间转换到HSV空间,保持色调分量不变,采用中心可自适应调节的双曲正切函数增强图像的全局亮度,局部细节非线性变换处理进一步提高图像的局部对比度,运用线性拉伸对饱和度进行调整,实现颜色补偿。实验结果表明该方法去雾效果显著,且颜色自然。实验还结合方差、熵和算法运算时间等参数,对该算法与多尺度Retinex算法作比较,验证了该算法在图像对比度、细节增强方面的优越性,且算法速度快,具有应用于实时图像处理的能力。

Image enhancement is important for intelligence management system.It has potential applications in many fields such as traffic management,highway toll collection station,ship,airport and etc.The adjustment ability of logarithmic function,hyperbolic tangent function and inverse hyperbolic tangent function for fog-degraded images were compared.It was proved that hyperbolic tangent function had better ability to regulate brightness than logarithmic function.Accordingly,an image enhancement algorithm based on single scale Retinex was proposed.The algorithm conversed RGB color space to HSV space.Hue was kept unchanged.The hyperbolic tangent function based on center self-adaptive adjustment was used to enhance overall brightness of image.Local non-linear transformation was adopted to improve local contrast of images.And linear stretching was used to adjust saturation.Then color compensations are achieved.Experiments show that this kind of algorithm has significant effects on defogging and making color full of nature.Combined with parameters such as variance,entropy and arithmetic,experiments were conducted to make comparisons for algorithms based on single scale Retinex and multi-scale Retinex.It is proved that algorithm based on single scale Retinex has more advantages on image contrast and detail enhancement.And it has faster arithmetic speed.Generally,the algorithm based on single scale Retinex can be applied in real-time image processing.