Introducing object reflectance property and sensor spectral response into empirical mode decomposition based MODIS and TM image fusion

Abstract

In existing image fusion methods, there is little work concerning the fusion of Moderate Resolution Imaging Spectroradiometer (MODIS) images with Landsat Thematic Mapper (TM) images. In most cases, the object reflectance property and sensor spectral response are not considered, which produces some undesirable effects such as decreased resolution over injection images and slightly modified spectral signatures in some features. Starting from a simplified land surface reflection model, we deduce a general method that takes both aspects into account for injecting features from the TM image into the MODIS images, trying to preserve spectral signatures of the latter and improve the spatial resolution to that of the former. This method is further improved using empirical mode decomposition (EMD) by considering the difference between the detail radiation absent from the MODIS images and that appearing in the TM image. In the experiment, visual inspection and Wald’s protocol are used to assess the qualities of the fused MODIS images qualitatively and quantitatively, respectively. Compared to many existing state-of-the-art fusion methods, the proposed method produces fused MODIS images that are closer to the image the corresponding virtual MODIS sensor would observe if it worked at the spatial resolution of the TM image. Extensive assessment results demonstrate that the proposed method is encouraging for increasing spatial details of the MODIS image with its spectral properties reliably preserved. The proposed method is recommended as a useful tool for the fusion of images with significantly different spectral and spatial resolutions.