 The study proposes an improved spatial spectral classification method using minimum spanning forest, MSF, algorithm for hyperspectral images in urban areas. Two spatial texture features are first extracted using wavelet and GABA filters, then the weighted genetic, WG, algorithm is used to obtain the subspace of hyperspectral data and texture features. These are fed into a novel marker-based MSF classification algorithm that achieves approximately 17% and 14% better overall accuracy than the original MSF-based algorithm for Pavia University and Berlin Image Datasets, respectively. This article was authored by Davoud Aghbari.