室外自然场景下的雾天模拟生成算法

摘要

In this paper, a new algorithm is proposed to simulate the foggy scene based on a real image. The algorithm takes three steps. Firstly, it segments the input image into several semantic regions, which are labeled as the sky, the ground and the targets on the ground. Secondly, it estimates depths for different regions. Finally, it synthesizes virtual fogs in the scene using an atmospheric scattering model. The theory and experiment show that the computational complexity of the proposed algorithm only relates to the resolution of the image, and the synthetic foggy scenes are visually realistic.%为了通过单幅真实拍摄的图像生成一系列具有雾天效果的虚拟场景,提出了一种完全基于图像的室外自然场景的雾天模拟算法,该算法首先对图像内容进行语义分割,将不同的场景内容标签为天空、地面和立于地面上的物体等；其次针对不同类型的场景分别进行深度信息的解析；最后利用大气散射模型进行雾天的模拟与仿真,从而得到室外自然场景的雾天虚拟图像.理论分析和实验结果表明,该算法的效率仅与图像的分辨率有关,且生成的雾天虚拟场景具有视觉真实感.