Multispectral image classification based on improved weighted MRF Bayesian.

This paper presents a novel nonparametric supervised spectral-spatial classification method for multispectral image. In multispectral images, if an unknown pixel shows similar digital number (DN) vectors as pixels in the training class, it will obtain higher posterior probability when assuming DN vectors of different classes follow a certain type of statistical distribution. According to statistical characteristics about DN vectors, the proposed method assumes the vectors follow a Gaussian mixture distribution in each class. Particularly, adaptively Bayesian nonparametric method is developed to estimate the optimal settings in distribution model appropriately. Then, we construct an anisotropic hierarchical logistic spatial prior to capture the spatial contextual information provided by multispectral image. Finally, optimized simulated annealing algorithm is conducted to estimate the maximum a posteriori. The proposed approach is compared with state-of-the-arts classification methods of multispectral images. The comparison results suggested that the proposed approach outperformed in overall accuracy and kappa coefficient. A novel multispectral image classification scheme is proposed in this paper.GMM is introduced to describe the statistical properties of each class.Anisotropic MLL is constructed to capture the spatial contextual information.The proposed Bayesian approach combines GMM and anisotropic MLL spatial prior.