偏振成像可以获得比传统成像多一维的偏振信息，在对透明、高反光和纯黑色等目标的探测中具有独特的优势。偏振相机采用覆盖偏振滤波阵列的图像传感器结构，每个像元只能获取一个方向的偏振态信息，需要去马赛克算法处理得到完整的偏振图像。基于插值的偏振图像去马赛克方法具有实时性高和应用性强的潜能，然而以往的插值方法在边缘处容易出现伪影问题。对此，本文提出了一种基于边缘感知引导滤波的偏振图像去马赛克算法。算法的关键是一个新的边缘感知引导滤波：首先，加入一个偏振图像边缘探测器生成边缘敏感的引导图；然后，对引导过程用梯度域引导滤波进行改进，使边缘和平面图像块具有不同的引导权值。应用改进的边缘感知引导滤波对插值图像进行优化，得到了保持边缘的结果，减少了边缘伪影。本文在单色和彩色偏振图像数据集上同几种先进算法进行了实验比较，结果表明提出的算法为偏振成像提供了一种快速有效的解决方案。

Polarization imaging can obtain an additional dimension of polarization information compared to traditional imaging, and has unique advantages in detecting transparent, highly reflective, and pure black targets. The polarization camera adopts an image sensor equipped with a polarization filtering array, where each pixel can detect only one channel of raw image data, so it requires demosaicking processing to obtain high-quality polarization images. Those polarization image demosaicking methods based on interpolation have the potential for high real-time performance and strong applicability. However, previous interpolation methods are prone to artifacts at the edges. In this paper, an improved polarization image demosaicking algorithm based on edge-aware guided filtering is proposed. The key to the algorithm is a new edge-aware guided filter: first, a polarization image edge detector is used to generate better guidance images; Then, the guidance filtering is replaced with gradient domain guided filtering, so that the edges and flat image blocks have different guidance weights. Then this edge-aware guided filtering is applied to optimize the interpolation images, which preserves image edges better and reduces edge artifacts. Comparing with several advanced algorithms on monochrome and color polarization image datasets, the results show that the proposed algorithm is a fast and effective solution for polarization imaging.