由于硬件技术和发射成本的限制，卫星图像的空间分辨率和时间分辨率之间存在相互制约的问题。为了获得兼具高空间和高时间分辨率的数据，遥感图像的时空融合技术应运而生。近年来，卷积神经网络也被成功应用于该领域中，产生了一些效果良好的时空融合方法。但是这些方法需要较多的图像样本对，限制了它们的应用。针对此问题，本文提出了一种单样本对卷积神经网络时空融合方法（SS-CNN）。该方法以高空间分辨率图像的波段平均图像提供的空间信息激励卷积神经网络建立高、低空间分辨率图像间的超分关系，进而利用该超分关系映射求解目标高空间分辨率图像。在实验中使用两个模拟数据集和一个真实数据集对该方法进行了测试，实验结果表明了其有效性。

Due to the limitations of hardware technology and launching cost, there is a tradeoff between the spatial resolution and temporal resolution of satellite images. In order to access to the data with both high spatial and high temporal resolution, spatio-temporal fusion (STF) of remotely sensed images came into being. In recent years, convolutional neural networks (CNNs) have been successfully adopted in this field and some efficient STF methods based on CNNs were developed. However, these methods require a significant number of training image pairs, where each pair generally consists of a high spatial resolution image and a low spatial resolution image. Such a requirement limits the applicability of STF methods to real scenarios, as in many cases there is no wide availability of image pairs for training. To overcome this important limitation, in this paper we introduce a single image pair-based method (based on CNNs) for STF of remotely sensed images. Our method, called SS-CNN, uses the spatial information provided by the average image (obtained across the available spectral bands) of the high spatial resolution image to perform CNN-based super-resolution mapping (SRM) between the low and high spatial resolution images. Three experiments, including two simulated and one real ones, were used to evaluate the STF accuracy of SS-CNN. The obtained experimental results clearly demonstrate the effectiveness of our newly proposed method.