中文摘要：

雾霾限制了户外图像的可见度并降低了图像的对比度.去雾是基于有雾图像成像模型,通过恢复场景辐射度去除雾的干扰,得到可能的清晰图像.现有的图像去雾方法依据所采用先验信息的不同而大体分为局部和非局部两类.Berman等人基于清晰图像在RGB空间的非局部聚类特性,构造了有雾图像每一颜色类的几何表示-雾线（Haze-Line）.雾线的最大辐射坐标（LRC：Largest Radial Coordinate）是估计初始透射率的关键.从统计学的观点出发,给出了LRC的一个无偏估计.实验结果表明,该方法可得到至少与原方法相当的结果.

英文摘要：

Haze limits visibility and reduces the contrast ratio of out-door images.Based on the common haze image formation model,image dehazing can get possible clear image by recovering the scene radiance which aims to remove the interference of the fog.According to the distinct prior information,the existing image haze removal methods can be divided into local and non-local.Berman and others construct a geometric representation（haze-line）for each color class of haze image which is based on the non-local clustering characteristics of clear image in the RGB space.The largest radial coordinate（LRC）of haze-line is the key to estimate the initial transmission.From statistical point of view,this paper proposes an unbiased estimation of LRC.The experiments show that the proposed method can gain at least comparable results with original method.