An elementary fact in our system is the whole set of available information features for each pixel. Some of them may not be immediately available but can be computed. Note that at this step we suppose that a supervised classification has already been done and for each pixel we keep all possible choices with their various certainty factors.

So, one fact appears as follows, where x,y are the pixel coordinates:

Class	Certainty factor
1
2
3
......	......
9

Successive application of knowledge base rules will update the certainty factors , ... When no more "pointwise" rule is eligible, a conclusion can be made. We will consider here that we have reached the decision step, so the pixel will be labeled with the class whose certainty factor is the highest one.

Note: If we decided to introduce "structural rules" the preceding decision step would be skipped and a new updating process performed.