Single Sensor Image Fusion Using A Deep Convolutional Generative Adversarial Network

Recently deployed multispectral sensors can acquire multispectral images where different bands have different spatial resolution depending on wavelength. An example is the Sentinel-2 constellation which can acquire multispectral bands of 10 m, 20 m, and 60 m resolution, covering the visible, near-infrared (NIR) and short-wave infrared (SWIR) parts of the electromagnetic spectrum. In this paper, a method to perform image fusion of the fine and coarse spatial resolution bands to increase the resolution of the coarser bands is proposed. The method is based on a so-called Generative Adversarial Network (GAN) and uses a deep convolutional design for both the generator and the discriminator. In experiments, it is demonstrated that the proposed method gives good results when compared to state-of-the-art single sensor image fusion methods using both simulated and real Sentinel-2 datasets.