基于双边滤波的实时图像去雾技术研究

Real Time Image Haze Removal Using Bilateral Filter

为了消除尘雾对室外机器视觉系统的影响,提出了一种基于双边滤波器的实时图像去雾方法.根据暗原色先验统计原理估计大气光强度,使用快速双边滤波器估计大气光幕,通过求解带雾图像成像物理方程恢复理想光照条件下的大气辐射强度,实现图像去雾.实验结果表明,该方法对彩色图像和灰度图像均可实现理想的去雾效果,对于(320×240)像素的图像,可达到20帧/s以上的处理速度,满足视频图像实时处理要求.

A real time image dehazing method using bilateral filter is proposed to eliminate ill effects of haze or smoke for outdoor machine vision system. Airlight vector was estimated based on statistics of dark channel prior at first, and then the atmospheric veil was inferred by the estimation of fast bilateral filter. The ideal scene radiance could be recovered by solve imaging physical equation of hazing scene. Experiment results indicate that this method performs well both for color and gray images with haze. Its processing speed is faster than 20 frame/s dealing with images of 320 × 240 pixels and could satisfies the demands of real time image applications.