摘要
本文提出了一种单幅雾天图像分割复原的方法,首先利用黑体原理估计粗糙的大气光传输图,然后使用拉普拉斯修补矩阵对分割之后的传输图进行修补,当场景目标和大气光很类似时,采用置信传播推断算法纠正传输图,经过修补和推断之后的传输图能够准确反映光线通过雾的传输过程,结合雾天图像光学模型,从雾的物理特性上去除雾对图像的影响.实验结果表明,本文提出的雾天图像分割复原算法能够有效恢复出清晰图像,并能获取相应雾天图像深度信息.
In this paper,we propose a method to restore a single haze image using segmentation algorithm.Firstly,the dark channel prior is used to estimate the rough transmission.Using laplacian matting matrix,we can matting the transmission that is segmented.When the scene objects are inherently similar to the atmospheric light,the transmission map is rectified by belief propagation algorithm.Using the transmission map that is repaired and rectified and the haze image optical model,we can physically recover the scene radiance.The experimental results demonstrate the new method abilities to remove the haze layer as well as provide a reliable depth map.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2010年第10期2279-2284,共6页
Acta Electronica Sinica
基金
国家自然科学基金(No.60705015,No.60805019)
关键词
图像复原
分割修补
去雾
暗原色
image restoration
segment matting
dehaze/defog
dark channel prior