面向内河雾天图像的大气光亮度值估算方法研究  

Atmospheric Light Estimating Algorithm Based on Inland Haze Image

作　　者：黄明晶[1] 刘清[1] 熊燕帆[1] 郭建明[1] 

机构地区：[1]武汉理工大学自动化学院,武汉430070

出　　处：《交通信息与安全》2013年第3期33-38,共6页Journal of Transport Information and Safety

基　　金：国家自然科学基金项目(批准号:51279152)资助

摘　　要：在对内河雾天图像进行去雾复原的过程中,大气光亮度值是复原图像的关键参数之一,如何有效地选取天空区域来估算大气光亮度值是内河雾天图像复原的重点。通过对目前不同大气光亮度值的估算方法进行研究分析,提出利用K均值聚类方法分割天空区域,自动确定天空的聚类中心,将天空区域的亮度均值与雾最浓中心亮度值加权作为最终获得的大气光亮度值。采用该方法来提取内河雾天图像的天空并估算大气光亮度值的实验结果表明能有效地自动提取天空区域,从而更加准确地获得大气光亮度值,是一种适合内河雾天图像的大气光亮度值估算方法。In the inland image haze removal process,the atmospheric luminance value is one of the critical parameters of image restoration.how to effectively extract sky region to estimate atmospheric luminance value is the focus of the inland image restoration.In this paper,through studies and analysis on the current estimation methods for different atmospheric luminance values,K-means clustering method is proposed to split the sky area,automatically determine the cluster center of the sky and finally obtain atmospheric luminance value by weighting the luminance mean and the most concentrated center luminance value of sky area.Experimental results show that the proposed method can effectively extract the sky area;therefore,it can obtain more accurate atmospheric luminance value and it is suitable for atmospheric light estimation for inland haze image.

关 键 词：内河 K均值 大气光 亮度值 大气模型 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]