Adding spatial information for texture segmentation

Hi, I found an example of texture segmentation in Matlab Help-files which adds a spatial map with the data. See the Matlab Example. My question: A. How does it actually help the classifier? and B. Could this technique be used with a classifier/clustering algorithm other than kmeans, i.e. how about hierarchical clustering? C. Please suggest clustering algorithms that works on data like the one shown in this example. D. How can I extract which Gabor-filter had the most significant contribution in the clustering process?