摘要
针对暗通道先验理论去雾算法景深突变区域容易出现光晕、不满足暗通道先验理论区域出现颜色失真、大气光估计不准确容易导致偏色现象,提出了自适应透射率和大气光估计的去雾算法。通过Canny算子结合形态学膨胀运算确定易出现光晕效应的区域,在该区域上采用像素等级差诱导高斯核函数得到的自适应高斯滤波器来消除光晕效应。其次,提出一种天空区域透射率估计方法处理因不满足暗原色原理出现的天空区域颜色失真问题。最后,提出先估计大气光方向,再最大化大气光方向与前1%最亮像素点的相似度的大气光估计方法。实验表明,该算法获得的透射率图精细,复原图像自然,能够有效去除光晕效应,修正天空区域的颜色失真。
In haze removal based on dark channel prior theory, the main failure includes： 1） halo effect in the area with depth mutation; 2） color distortion in the field where the dark channel prior theory is not satisfied; 3） color cast due to inaccurate estimation of airlight. In this paper, an adaptive transmission and airlight estimating method was proposed. Canny operator and dilation operation were applied in hallo-effect area detection; on the other hand, an adaptive Gaussian filter based on pixel-level Gaussian kernel function was used to eliminate the halo effect. A method of estimating sky transmission was applied to deal with the color distortion because of dissatisfaction of dark channel prior. Finally, a new airlight estimating method was proposed： first, the direction of airlight was estimated; second, the top 1% brightest pixels were picked; third,the pixel with the highest similarity to the estimated direction of airlight was regarded as airlight. The experiment results show that the proposed approach can achieve refined transmission image, unaffected recovery pictures. Besides, the proposed method can effectively eliminate the halo effect, revise the color cast of sky region.
出处
《计算机应用》
CSCD
北大核心
2017年第A01期176-179,186,共5页
journal of Computer Applications
基金
中央高校基本科研业务费资助项目(CQDXWL-2013-Z009)
关键词
暗通道图
透射率
大气光估计
大气散射模型
自适应高斯滤波器
dark channel image
transmission
arilight estimation
atmosphere scattering model
adaptive Gaussian filter
