An Overview on Pixel-Level Image Fusion in Remote Sensing

Pixel-level image fusion is an important part of image fusion algorithms which can combine spectral information of coarse resolution imagery with finer spatial resolution imagery. The objective of this paper is to present an overview of pixel-level image fusion algorithms used for effective information interpretation of remotely sensed imagery of various spatial and spectral characteristics. According to their different characteristics, these algorithms are categorized into four types, i.e., color space model algorithms, statistical/numerical algorithms, multiresolution decomposition algorithms and radiometric/spectral algorithms. They are investigated respectively and comparative analysis on the performance of these algorithms is conducted on a set of multispectral(MS) images and a Systeme Pour l'Observation de la Terre (SPOT) Panchromatic(PAN) image of the same scene. The effectiveness of these algorithms is evaluated quantitatively and qualitatively. Experimental results indicate that multiresolution decomposition based algorithms especially the discrete wavelet transform shows a comparative better performance on the test data than the other three types of fusion algorithms. However the suitable selection of a proper pixel-level fusion algorithm depends on the merits of each method, relevant applied situations and the characteristics of the source data.