大气中水分子及微小颗粒对光的散射和吸收, 使得在雾天条件下获取的图像严重降质, 本文提出一种结合自适应雾气估计的快速单幅图像去雾算法。首先, 该算法从大气散射模型出发, 通过分析景深与亮度分量之间存在的相关关系, 提出线性系数利用亮度分量来近似估计出景深, 并通过最小滤波对明亮区域进行修正, 得到粗略透射率; 其次, 观察到散射系数值与雾浓度呈正相关, 从而结合雾浓度模型与指数函数提出自适应散射系数概念, 估计出较准确的透射率; 最后, 根据大气散射模型复原出无雾图像。实验结果表明本文算法可以复原出清晰自然的无雾图像, 明显提高了图像可见度, 且具有较低的时间复杂度。

Owing to the scattering and absorption of light by water molecules and other tiny particles in the atmosphere, images captured in hazy conditions are inevitably seriously degraded. To address this problem, a fast single image dehazing algorithm combined with adaptive haze estimation was proposed in this paper. The algorithm begined by considering the atmospheric scattering model and assumed a correlation between scene depth and luminance components. Subsequently, a linear coefficient was proposed to approximate the scene depth using the luminance components, and overly bright regions were corrected via minimum filtering to obtain a coarse transmission. As the value of the scattering coefficient was related to the haze concentration, the concept of an adaptive scattering coefficient was proposed by combining the haze concentration model with the exponential function to estimate the accurate transmission. Finally, a haze-free image was restored via the atmospheric scattering model. Experimental results demonstrated that the proposed algorithm could recover a clear, natural, and haze-free image, which significantly improves image visibility and has lower computational complexity than existing methods.