Sun glint in hyperspectral images (HSIs) leads to undesirable spectral interference, which severely affects subsequent image interpretation, such as environmental monitoring of coastal areas. Sun glint removal methods aim to recover a high quality image without sun glint from the original image. Most methods depend on an assumption that the near infrared band is strongly absorbed by water. However, this assumption is not always reliable because the infrared radiation in shallow or turbid water can be reflected back by the seabed or sediment, rather than being fully absorbed. Therefore, the reflected infrared radiation still contains sun glint and these methods cannot sufficiently remove sun glint from HSIs. To address this problem, a texture-aware total variation (TATV)-based method is proposed to remove sun glint from HSIs. The original HSI first is formulated as a desired clean image and a sun glint image. Then, in order to remove the sun glint, we propose a variational model where the different spectral characteristics of sun glint and other surrounding materials are considered. Specifically, we propose a texture-aware total variation regularized method to heavily penalize the variation of the sun glint areas. Experiments performed on simulated and real data sets demonstrate that our method can greatly outperform other state-of-the-art methods in removing sun glint.

高光谱图像（HSI）中的太阳眩光会导致不良的光谱干扰，从而严重影响后续的图像解释，例如沿海地区的环境监测。去除太阳光的方法旨在恢复高质量图像，而原始图像中没有太阳光。大多数方法取决于一个假设，即近红外波段会被水强烈吸收。但是，这种假设并不总是可靠的，因为浅水或浑浊水中的红外辐射可能会被海床或沉积物反射回去，而不是被完全吸收。因此，反射的红外辐射仍然包含阳光闪烁，并且这些方法不能从HSI中充分去除阳光闪烁。为了解决这个问题，提出了一种基于纹理感知的总变化量（TATV）的方法，以消除HSI中的太阳眩光。首先将原始HSI公式化为所需的干净图像和阳光闪烁图像。然后，为了消除太阳闪烁，我们提出了一个变分模型，其中考虑了太阳闪烁和其他周围材料的不同光谱特征。具体来说，我们提出了一种纹理感知的总变化正则化方法，以严厉惩罚太阳闪烁区域的变化。在模拟和真实数据集上进行的实验表明，在消除太阳眩光方面，我们的方法可以大大胜过其他最新方法。我们提出了一种纹理感知的总变化正则化方法，以严厉惩罚阳光闪烁区域的变化。在模拟和真实数据集上进行的实验表明，在消除太阳眩光方面，我们的方法可以大大胜过其他最新方法。我们提出了一种纹理感知的总变化正则化方法，以严厉惩罚阳光闪烁区域的变化。在模拟和真实数据集上进行的实验表明，在消除太阳眩光方面，我们的方法可以大大胜过其他最新方法。