Pansharpening based on details injection model and online sparse dictionary learning

Abstract

Pansharpening is a process of combining a high-resolution Panchromatic (PAN) image with a low-resolution Multispectral (MS) image to obtain a high-resolution MS image. Pansharpening can not only overcome the physical and technical limitations of satellite sensors, but also improve the quality of images and obtain a more detailed description of the scene. In this paper, a novel method of Pansharpening is proposed which adopts the ARSIS concept. The spatial information missing in the low-resolution MS image can be extracted from the PAN image. This is achieved by image decomposition based on sparse representation. Then, the information is injected into the MS bands by the details injection model. In the dictionary training, we adapt Online Sparse Dictionary Learning (OSDL), which can shorten the dictionary training time and improve the fusion image effect. Experimental results show that the fusion image of the proposed method has stronger spectral performance and more detailed spatial information.