基于改进大气散射模型的图像去雾算法

Image dehazing algorithm based on improved atmospheric scattering model

由于雨雾天气的影响,清晰图像的获得较为困难,通常存在能见度低、对比度差、细节信息缺失等问题。针对上述问题,文中提出一种鲁棒性高的图像去雾算法。首先,将输入图像转换为细节图像,衰减图像并重新定义三个颜色通道,根据最小颜色损失原则对颜色进行补偿并平衡三个颜色通道的差异;其次,通过改进的大气散射模型EASM和暗通道先验算法解决图像发暗的问题,去雾结果明显、颜色鲜艳、细节清晰。在自然图像和合成图像数据集上进行对比实验并设计消融实验,结果表明,所提算法在信息熵、FADE、自然图像质量评估器(NIQE)、结构相似性(SSIM)等方面表现优于最新的去雾算法,具有较高的鲁棒性和应用前景。

Low visibility,poor contrast and detail information missing will occur to the images due to rainy and foggy weather,so it is difficult to obtain a clear image.In view of this,a robust image dehazing algorithm is proposed.The input image is converted into a detailed image,and then the detailed image is attenuated and the three color channels are redefined.According to the principle of minimum color loss,the color is compensated and the differences among the three color channels are minimized.The image darkening is eliminated by the enhanced atmospheric scattering model(EASM)and dark channel prior(DCP)algorithm.The dehazing results are obvious,with bright colors and clear details.Comparison experiments are performed on natural image dataset and synthetic image dataset and ablation experiments are designed.The results show that the proposed algorithm outperforms the latest dehazing algorithms in terms of information entropy,FADE,NIQE(natural image quality evaluator)and SSIM(structural similarity index measure).Therefore,the proposed algorithm has high robustness and broaden application prospects.