Context-Based Scene Recognition Using Bayesian Networks with Scale-Invariant Feature Transform