期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2006
卷号：XXXVI Part 7
出版社：Copernicus Publications
摘要：Remotely sensed images are invaluable to acquire geospatial information about earth surface for the assessment of land resources and environment monitoring. In most cases, the information provided by a single sensor is not complete or sufficient. Therefore, images collected by different sensors are combined to obtain complementary information. Each remote sensing sensor has its own advantage and disadvantage over the other sensors. SAR (Synthetic Aperture Radar) sensors are active sensors and can collect images during day and night without being affected by weather conditions. SAR sensors are capable of sensing the geometry and structure of the features such as terrain topography, thickness and roughness of surface cover. They also sense the moisture content and presence of vegetation. However, Visible-Infrared (VIR) sensors are passive sensors that sense the electromagnetic energy reflected from surface. Therefore, the information provided by the SAR data alone may not be satisfactory for a detailed analysis of the terrain, since it does not has the capability of collecting spectral information about terrain cover types. For this reason, fusion of VIR and SAR images provides complementary data to increase the amount of information that can be extracted from the individual input images. For an optimal image fusion, some criteria should be defined for algorithmic development. The success of the fusion strongly depends on the criteria selected. In this work, a pixel based image fusion algorithm is proposed. The new method forms the fused images as the linear combination of the input images. The method employs adaptive windows to establish statistical relationships between the input images to calculate new fused pixels. The fused pixels are calculated using two criteria: 1) Variance of the local window in fused image should be equal to the variance of the corresponding window in higher resolution image to transfer spatial detail. 2) Mean of the local window in the fused image should be equal to the mean of the window in the original lower resolution image to retain the color content. This paper describes the principles of the proposed approach and assesses its properties. To test the performance of the new approach, the images from different sensors are fused as well as images from same sensors using proposed algorithm and PCA, Brovey and Multiplicative image fusion methods. For the different type of sensors, SAR image is fused with Ikonos, Quickbird (QB) and Landsat Thematic Mapper (TM) XS images. For the same sensor type, QB pan image is fused with QB XS and Ikonos XS images. The results are evaluated visually and analytical derivation and graphic results are presented
