为了解决高动态红外图像在常规显示设备上显示时容易出现图像整体对比度低、弱小目标细节模糊等问题，提出了一种基于引导滤波图像分层的红外图像细节增强算法，并从算法理论分析和仿真结果两方面验证了引导滤波具有更好的边缘保持能力，能有效避免增强后出现“伪边缘”的缺陷。另外，针对原始全局的引导滤波算法对整幅图像各个区域使用相同的规整化因子，容易产生“光晕”现象的缺陷，本文在局部方差加权引导滤波算法的思想上，提出了基于 LoG边缘算子的加权引导滤波算法。实验结果表明本文算法具有良好的细节增强效果，特别是对图像中的弱小目标；另外，相比目前应用广泛的双边滤波算法，本文算法运行时间要快得多，具有实时处理的应用前景。

To solve the problem that high dynamic infrared images are prone to appear with low image contrast and weak target fuzzy details when displayed on a conventional display device, a novel infrared image detail enhancement algorithm based on the hierarchical processing by a guided image filter is proposed. Furthermore, this study has verified that the guided filter has improved ability of edge keeping and can effectively avoid the “pseudo edge” using the theoretical analysis and simulation results. The original global guide filter algorithm uses the same neat factor among the whole image regions, which makes it is prone to exhibit halos near some edges. Based on the idea of the local variance weighted guided filter, an improved locally weighted guided filter using the Laplacian of Gaussian (LoG) operator is proposed. The experiments in this study show that this algorithm has good detail enhancement effect, especially for small targets in the images. Additionally, compared to the widely used bilateral filtering algorithm, this algorithm runs considerably faster, and has potential applications in real-time processing.