Fusion of satellite images in urban area: Assessing the quality of resulting images

The fusion of panchromatic and multispectral satellite images is an important issue in many remote sensing applications, especially in urban area. The popular image fusion methods in remote sensing community usually distort the spectral characteristics. To reduce the spectral distortion, some image fusion techniques have been developed. This paper addresses the issue in quality assessment of fused images from three recently developed methods. These are synthetic variable ratio (SVR), smoothing filter-based intensity modulation (SFIM) and Gram_Schimdt transform (GS). We employed these methods in image fusion of Landsat 7 ETM+ panchromatic with multispectral images and Quickbird panchromatic with multispectral images. The quantitative methods such as standard deviation, information entropy, correlation coefficient, and spectral bias index were used to assess the quality of fused images. The results indicate that different approaches have their specific properties and adapt to different purposes based on spectral fidelity and high spatial frequency information gain. The quality of fused images based on SFIM and SVR methods is better than that of GS method, respectively, in mediumresolution images and high-resolution images in urban area. Therefore, the SFIM and SVR methods can meet the needs of mapping-oriented fusion, classification-oriented fusion, and visualization-oriented fusion purposes, respectively in mediumresolution images and high-resolution images in urban area.