新的雾霾退化场景再现快速算法

Novel fast rendition algorithm for dehazed image

雾霾天气条件下,由于大气粒子的散射作用,使采集图像对比度、清晰度等方面产生降质。针对这一问题,提出了一种新的图像雾霾去除算法。算法从单色大气散射模型出发,根据有关大气光衰减项的先验知识与假设,构建有约束最优化问题对大气光衰减项进行直接求解。根据散射模型与大气光衰减项求解结果实现对原始场景反照率的恢复。实验证明,所提算法能够较好地对具有不同景深的场景图像信息实现恢复,提升场景视见度,算法鲁棒性较好,与同类算法相比运行效率提高1倍以上,能够较好地运用于智能交通监控等可见光计算机视觉系统。

Collected images are often degraded in terms of contrast and visibility by scattering caused by atmospheric particles. To solve the problem, a new fast rendition algorithm for haze degraded image was proposed. The algorithm was based on the monochrome atmospheric scattering model. According to the prior knowledge and the priori assumptions of the airlight, the constrainted optimal problem was constructed to estimate the airlight. And the scene albedo was restored with scattering model and the airlight. The experimental results show that the proposed algorithm achieves great restoration of images with various depth maps and improves the visibility of the scene. Besides, the algorithm is robust. Compared with other algorithms, the running efficiency of the algorithm is more than doubled. And the algorithm can be implemented in intelligent traffic monitoring and some other visible light computer vision systems.