雾天降质图像的快速复原  

Fast restoration of haze-degraded image

作　　者：吴笑天[1,2] 鲁剑锋[1] 贺柏根[1] 吴川[1,2] 朱明[1,2] 

机构地区：[1]中国科学院长春光学精密机械与物理研究所,吉林长春130033 [2]中国科学院长春光学精密机械与物理研究所航空光学成像与测量中国科学院重点实验室,吉林长春130033

出　　处：《中国光学》2013年第6期892-899,共8页Chinese Optics

基　　金：中国科学院航空光学成像与测量重点实验室开放基金资助项目(No.Y2HC1SR125)

摘　　要：针对在雨雾霾天气条件下,大气介质的散射和吸收作用导致光电成像系统接收的图像对比度降低,细节模糊不清及颜色偏移,提出通过快速图像复原来解决此类图像退化问题。基于大气成像光学模型,在暗通道先验的理论基础上,提出了一种基于形态学滤波器的快速估算暗通道图像的方法,并采用参数自适应调整方法来抑制暗通道先验不满足时的大片天空/白墙区域的颜色失真现象。实验结果表明,该算法能够有效快速复原雨雾天气条件下的降质图像,对于600×400大小的图像,其Matlab复原仿真时间仅为0.4 s,复原后的图像主观视觉质量明显提升,其大片天空/白墙区域的颜色失真得到有效抑制。Under haze , fog, and rain weather conditions , the image received by a photoelectric imaging sys-tem will lose the contrast and color fidelity owing to the absorption and scattering in atmosphere .To overcome the image degradation mentioned above , this paper proposes a new dark channel image estimation method u-sing morphology filter based on the analysis of the atmospheric optical model and the dark channel priority .It further utilizes a dynamic parameter strategy to solve the color distortion in a large area of sky or white wall . Results on a variety of outdoor haze images show that the proposed method can effectively restore a haze -de-graded image and improve the image quality .As for the image with 600 pixel ×400 pixel, the restoration sim-ulation time by Matlab is only 0.4 s.The dynamic parameter strategy is simple but effective to reduce the color distortion phenomenon .

关 键 词：降质图像 图像复原 暗通道先验 形态学滤波器 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]