摘要
遥感影像中薄云的存在为影像的判读带来了极大的影响,通过研究薄云对Landsat影像造成的影响,提出一种加权梯度融合变分模型。通过在无云区域采用较小权重以保持影像自身信息,薄云区域则采用较大权重将参考影像的梯度信息融入待修复影像,改进了梯度模型在无云区域过度增强细节而造成的失真。采用暗通道法和梯度融合法与该方法进行比较,实验结果表明:该方法在有效去除薄云的同时对无云区域有较好的保真效果。
The presence of thin cloud in remote sensing images has brought great impact for follow-up image interpretation.To remove thin cloud of Landsat images,a weighted gradient-based total variation fusion method is proposed.Considering the highly correlation and complementarity between the infrared and visible bands,we choose gradient information of infrared bands as reference.In our method,lower weights are used in cloudless areas to integrate the gradient information of reference images into the restored images.By doing so,the spectrum information of cloudless areas is well kept.Oppositely,higher weights in thin cloud areas are used to remove cloud.In contrast with dark channel method and gradient-based method,our method is effective visually to remove thin cloud and modify spectral distortion causing by excessive detail enhancement of gradient-based model.Quantitatively,the difference index R of our method in the cloudless areas is significantly lower than that of other methods.
出处
《遥感技术与应用》
CSCD
北大核心
2016年第3期511-517,共7页
Remote Sensing Technology and Application
关键词
薄云去除
梯度融合
变分模型
加权
Thin cloud removal
Gradient-based fusion
Total variation model
Weighted
