摘要
雾霾环境下,由于空气中悬浮颗粒等物质的散射作用导致探测器获取的图像质量下降.针对雾霾天气图像退化问题提出一种全局参数估计的颜色空间转换偏振去雾算法.首先,通过求取表征偏振信息的强度图、偏振度图和偏振角图,将其映射到HSI颜色空间增强目标特征信息,再将HSI颜色空间转换到RGB空间;其次,通过导向滤波优化暗原色先验原理求取的透射率图,运用形态学方法重建图像求取局部极大值自动估算无穷远处大气光值;接着结合优化后的透射率图估算大气光偏振度信息,并通过归一化互信息对大气光偏振度进行优化;最后通过偏振去雾模型恢复雾霾导致的退化图像.通过信息熵、平均梯度和灰度方差三个指标结果可以看出此算法要优于其他的算法,解决了人工估算参数产生误差的问题,能够明显提升雾霾图像的质量.
In the foggy environment,the quality of the image obtained by the detector is decreased due to the scattering of suspended particles and other substances in the air.Aiming at the problem of image degradation in the foggy environment,a color space conversion polarization defogging method is proposed based on global parameter estimation.Firstly,the intensity,polarization degree and polarization angle maps representing the polarization information are obtained,which are mapped to the HSI color space to enhance the target feature information,and then the HSI color space is converted to RGB space.Secondly,the transmittance map obtained by dark channel prior color is optimized by guided filtering,and the local maximum value of the image is reconstructed by the morphological method to estimate automatically the atmospheric light at infinity.Then,the polarization information of atmospheric light is estimated by combining the optimized transmittance map,and the polarization degree of atmospheric light is optimized by normalized mutual information.Finally,the degraded image due to the weather conditions such as haze is recovered by the polarization defog model.Through the information entropy,average gradient and gray variance,it is found that this algorithm is better than other algorithms,which solves the problem of error in manual estimation parameters and can significantly improve the quality of foggy image.
作者
李荣华
唐智超
李宏亮
朴俊峰
LI Ronghua;TANG Zhichao;LI Hongliang;PIAO Junfeng(School of Mechanical Engineering,Dalian Jiaotong University,Dalian 116028,China)
出处
《大连交通大学学报》
CAS
2022年第3期66-71,共6页
Journal of Dalian Jiaotong University
基金
辽宁省自然科学基金资助项目(20180520015)
辽宁省高等教育创新人才支持计划资助项目(LR2018050)。
关键词
颜色空间
暗通道先验
偏振去雾模型
形态学重建
互信息
color space
dark channel prior
polarization defogging model
morphological reconstruction
mutual information
