| Summary: Transferring spatial details of a high-resolution image into a low-resolution multispectral image is called image fusion. Different fusion methods produce different quality fused images. Some of these methods may damage to the spectral quality of low-resolution multispectral image while transferring spatial details into it. In literature, there are some metrics that are used to evaluate the spectral and spatial quality of fused images. These metrics have some disadvantages so they may not be successfull to examine the spectral quality of fused images. The main goal of this study is to present the spectral qualities of fused images using post-classification accuracy analysis. In order to do this, Worldview-2, Landsat ETM+ and Ikonos multispectral images are fused with their own panchromatic bands and another Ikonos image is fused with Quickbird pan-sharpened image by using 11 different fusion methods. These methods are IHS, CN, HPF, PCA, Multiplicative, Ehlers, Brovey, Wavelet, Gram-Schmidt, Criteria Based Fusion Method and a novel method in literature. Some metrics that are used to evaluate the quality of fused images are performed. Fused images are then classified with 6 different classification methods by using exactly the same signatures. Minimum Distance, Binary Encoding, Support Vector Machines, Random Forest, Maximum Likelihood and Artificial Neural Network classification methods are used. Examining the classification results and metrics together, HPF image fusion method is found out to be the most successful one in preserving the spectral quality of the data sets used while increasing the spatial resolution.
|
