Geographic information of urban land-cover types is important for urban planning. This study explored the use of Landsat ETM imagery for mapping urban land-cover types in Ho Chi Minh City (HCMC) in 2010. Spectral unmixing model estimates abundance fractions of surface targets at sub-pixel level was used for urban-land cover mapping. This model was trained using endmembers extracted from the original image using minimum noise fraction (MNF) method. The mapping results was assessed using ground-verification data. The comparison results between classification map and ground-truth data revealed the overall accuracy of 89.6% and Kappa coefficient of 0.86. The results achieved from this study could be useful to assist local authorities in urban planning.