Naive Bayes Classifier Biased Output?

I'm using Emgu CV to implement a machine learning technique in c# to classify pixels of my image into 3 different categories.

Everything works perfect so far, but the problem is that it is fully automatic. I want to make it semi-automatic which mean the user can "give wei开发者_Go百科ght" to each of those 3 outcomes. This is to give the user ability to well-tune the outcome.

Any idea how?

The first thing I can think of is to actually modify the input in a way that it would have bias to one of the outputs (for example make it more red by modifying red channel). But I though maybe there is a generic way of doing this that I'm not aware of.

Thanks.

Usually you'd do that by adapting the prior probabilities in the classification rule (what you get from the gaussian distributions is the likelyhood), but it seems that the implementation in emgucv does not allow you to do that.