摘要
雾或霾天气下,大气粒子对光的散射作用造成光学图像细节弱化,严重影响了后续的图像分析与处理任务.现有去雾算法存在去雾后图像信息丢失、产生模糊、天空区域过增强等问题.本文从偏振视角与暗通道先验理论出发,提出了一种基于直接透射光梯度特征引导的目标偏振度估算算法,用于图像去雾.通过偏振图像获取场景与大气的偏振信息;再以暗通道先验算法估计的直接透射光强的梯度特征为引导,估算目标偏振度;将估算的目标偏振度转为大气光强,经过原理性约束与引导滤波,得到优化的大气光强,进一步求解去雾图像与优化的目标偏振度.定性实验表明:本文算法去雾后,图像具有良好的平滑度,且克服了现有去雾算法存在的可见性弱、去雾残留、天空区域过增强等问题;定量实验表明:本文算法既不会造成图像信息丢失,也不会产生过多的噪声或模糊.综合对比9种具有代表性的去雾算法,本文算法具有良好的细节恢复能力、图像熵提升能力以及色调还原能力.
In foggy or hazy weather,the scattering of light by atmospheric particles weaken the details of optical images,which affects the subsequent image analysis and processing tasks seriously.The existing dehazing algorithms have problems such as the loss of image information,blurring and excessive enhancement of the sky after dehazing.Starting from the perspective of polarization and dark channel prior theory,this article proposes a target polarization degree estimation algorithm using the gradient feature of the direct transmission light intensity as guidance for image dehazing.The polarization information of scene and atmosphere are obtained from polarized images.Then,guided by the gradient feature of the direct transmission light intensity which is estimated by dark channel prior algorithm,the target polarization degree is estimated.The estimated target polarization degree is converted into atmospheric light intensity,and the optimized atmospheric light intensity is obtained after theoretical constrainting and guided filtering atmospheric light intensity,then the optimized target polarization degree and image after dehazing are solved.Qualitative experiments show that the image dehazed by the proposed algorithm has good smoothness and overcomes the problems of existing dehazing algorithms,such as low visibility,dehazing residue and excessive enhancement of the sky.Quantitative experiments show that the proposed algorithm neither causes the loss of image information,nor generates excessive noise or blurs.The comparison with nine representative dehazing algorithms shows that our proposed algorithm has good ability of restoring details,improving image entropy,and enhancing the degree of tonal restoration.
作者
徐万春
张焱
张景华
凌峰
李顺
XU Wan-chun;ZHANG Yan;ZHANG Jing-hua;LING Feng;LI Shun(National Key Laboratory of Science and Technology on Automatic Target Recognition,College of Electronic Science and Technology,National University of Defense Technology,Changsha,Hunan 410073,China;Academy of Military Science,Beijing 100091,China;College of Electronic Science and Technology,National University of Defense Technology,Changsha,Hunan 410073,China)
出处
《电子学报》
EI
CAS
CSCD
北大核心
2024年第6期2011-2024,共14页
Acta Electronica Sinica
基金
国家自然科学基金(No.62075239)。
关键词
图像处理
图像去雾
偏振图像
暗通道先验
偏振度估算
image processing
image dehazing
polarization images
dark channel prior
polarization degree estimation
