Again, we assume spatial independence between the classification labels as in equation 2. However, taking , where are the adaptive global class proportion parameters, the prior on is now:

The global class proportions, , are the relative weighting of each of the distributions in the mixture. The prior on is non-informative (uniform) over the range , and . Using this in equation 1, the posterior becomes: