半监督卷积神经网络遥感图像融合

近几年随着深度学习的发展,学者们利用卷积神经网络实现遥感图像融合取得了不错的效果.由于没有高分辨率多光谱图像作为参考图像,所以一般基于深度学习的方法在退化图像上训练模型,然后用训练好的模型去预测高分辨率多光谱图像,但是退化图像的融合过程并不能完全反映原始图像的融合过程.为了改善融合性能,提出了一种半监督卷积神经网络遥感...

Semi-supervised convolutional neural network remote sensing image fusion

With the development of deep learning in recent years, remote sensing image fusion methods based on convolutional neuralnetwork were proposed and presented with good performance. Because there is no high-resolution multispectral image as a reference, theconvolutional neural network is trained in the degraded images. The trained network is used to predict high resolution multispectralimages. However, the fusion process of degraded images cannot reflect the fusion process of original images. In order to improve fusionperformance, a semi-supervised fusion method based on convolutional neural network is proposed. The same fusion network is trained inthe degraded image and the original image simultaneously. Because degraded image fusion has the corresponding reference image, thesupervised learning method is used to train the fusion network. Moreover, the spectral loss is added to preserve the spectral information.However, there is no high-resolution multispectral reference image in the original image fusion. Spectral degradation network and spatialdegradation network are designed to train the fusion network. The experimental results show that the proposed method is better than thecompared method in preserving the spectral and details.