Single Sensor Image Fusion Using a Deep Residual Network

摘要

Single sensor fusion is the fusion of two or more spectrally disjoint reflectance bands that have different spatial resolution and have been acquired by the same sensor. An example is the Sentinel-2 constellation which can acquire multispectral bands of 10 m, 20 m and 60 m resolution from the visible to short-wave infrared (SWIR) regions of the electromagnetic spectrum. In this paper, we present a method based on a deep residual convolutional network to fuse the fine and coarse spatial resolution bands to obtain finer spatial resolution versions of the coarse bands. The benefits of the residual design are primarily that the network converges faster and it allows for deeper networks. Also, it improves the spectral consistency of the fused image. Using a real Sentinel-2 dataset, it is demonstrated that the proposed method gives good results when compared to state-of-the-art single sensor image fusion methods.