单一图像多滤波联合快速去雾算法

提出一种多滤波联合单一图像快速去雾算法。首先基于大气散射光特性,计算雾天图像最小通道图,给出大气耗射函数初估计;利用双边滤波保边缘的平滑特性,对雾天图像灰度图进行平滑增强处理;再以此滤波图作为引导图,对大气散射图进行引导滤波,生成准确的能体现深度变化的大气散射图。为了正确估计全球大气光,采用循环四分图形法求解雾天图像最小通道图中最亮像素;在此基础上生成视觉上较真实的清晰无雾图像。比较实验表明算法去雾效果更完整、视见度更高、速度更快。

Fast image dehazing from a single image based on multiple filtering methods

This paper proposes a new fast dehazing method based on multiple filtering methods. We firstly estimate an initial atmosphere scattering light through its characteristics, then refine it by bilateral filtering and guided filtering to generate a new atmosphere light map which removes the abundant texture information and recovers the depth edge information. And, the correct atmosphere light is generated by a hierarchical searching method based on the quad-tree subdivision. Finally, we solve the scene radiance using the atmosphere attenuation model. Comparing experiments show that the proposed algorithm could get a dehazingeffect and is sufficiently fast for real-time applications.