In this paper, a complete procedure is proposed to analyze and classify the texture of an image based Bayesian network classifiers. We apply this procedure in the residential areas detection. A simple case of Bayesian network called naive Bayes classifier is used to learn the positive and negative samples and to infer about the unknown regions. In this paper, each texture feature vector is labeled using fuzzy c-mean clustering, and the learning in general Bayesian networks provide the beliefs about the dependency of the regions for residential areas and non-residential areas with the of texture feature. In the learning process, the proposed methodology also handles unknown cases that are not correctly classified using existing samples and will be taken as samples in next learning process. Experimental results of residential areas detection on panchromatic remote sensing images are presented to illustrate the merit and feasibility of the proposed method