基于单目深度估计的红外图像彩色化方法

Infrared Image Colorization Based on Monocular Depth Estimation

针对夜视红外图像深度信息估计问题,根据红外图像的特点,用地平线检测和最大类间方差法分割得到一幅具有深度信息的图像,然后运用透视学中饱和度、色调对比度与深度的关系,用色彩传递算法对红外图像做彩色化处理,得到一幅视觉上具有立体感的彩色红外图像。

For the problem of estimating the depth from the night vision infrared image, according to the characteristics of infrared images, we use maximum between-cluster variance and the horizon detection methods to get a depth map of an infrared image. Then according the relationship between saturation, hue, and the depth in perspective, we transfer colors to the infrared image and obtain a color infrared image with three-dimensional visual perception.