Image fusion of Ikonos pan and multispectral images for classification of the urban environment

The paper compares different data fusion techniques in order to choose an appropriate technique for accurate urban mapping. Availability of high spatial resolution Ikonos and Quickbird imagery has made accurate mapping of urban areas more feasible. Though Ikonos multispectral images have a good spatial resolution of 4 m, it is often desirable to have an increased spatial resolution for a more accurate mapping. Ikonos panchromatic and multispectral images of the City of Fredericton, New Brunswick, Canada were fused to obtain 1-m multispectral images. This paper evaluates the results of the IHS (Intensity-Hue-Saturation), PCA (Principal Component Analysis), wavelet addition and wavelet substitution, and ARSIS ("Amelioration de la Resolution Spatiale par Injection of Structures") methods both visually and statistically. A maximum likelihood classification (MLC) was carried out on the original 4-m as well as on the fused 1-m multispectral images. The classified maps obtained are analyzed visually and also compared using classification accuracies.