基于红外图像与可见光雾化图像之间的相似性, 通过比较红外图像的灰度均值和图像亮暗程度的灰度阈值, 实现图像亮暗的自动判别, 进而判断图像是否需要灰度反转。对原图或原图灰度反转图采用类似可见光去雾的暗原色先验算法——伪暗原色算法求出红外图像的粗糙透射率, 结合导向滤波实现粗糙透射率的细化, 依据大气散射物理模型得到最终增强的红外图像。实验结果表明, 采用伪暗原色红外图像增强的算法可以提高红外图像的对比度, 突出其细节, 提高其信噪比, 并且具有良好的视觉效果。

Based on the similarities between infrared images and visible hazy images, automatic identification of an image whether it is light or dark is achieved by comparing the average of the image’s gray level and the threshold determining the degree of dark, and then it is known whether the image needs gray scale inversion. Next, similar dark channel prior (DCP) method called pseudo dark channel prior is implemented on original image or inversed image to get the coarse transmission of the infrared image; refined transmission is got by using the guided image filtering; final enhanced infrared image is obtained according to the atmospheric scattering physical model. Experimental results show that the infrared image enhancement algorithm based on pseudo dark channel prior can improve the contrast of infrared images, highlight the details, improve signal to noise ratio, and have a good visual effect.