How to generate a confusion matrix?

Dear all,

I need to calculate the precision, sensitivity and Fscore for a segmentation algorithm which I used on a 200 image database. Basically I have to overlay the mask I obtained, which, I guess, it represents the predicted class, over the orginial image, which is the ground truth and the actual class, I assume. This overlaying is done by the confusion matrix or do I have to do it separately? If so, how? Also I read that in the confusion matrix the classes must be of the same type, my images are RGB and the mask is binary, so my wild guess is that it wouldn't work like this.

I don't know how to do this overlaying thing and classification for the confusion matrix. If you could provide some step-by-step guidance it would be great!

Thanks in advance