Classification based on spectral signatures or apriori probabilities of an area with remote sensing imagery of 50m resolution or better vs using texture as one of the support information. How significant is the change in classification accuracy observed?