If a Haar feature matches that part of the image. There are as I said maybe a thousand or so within a full profile. So if we go back to one of these visuals. For example, this is a good one to start with [pulls up the first stage in the Haar cascade with just three face images overlaid with black and white rectangles]. For example there’s a rectangular region. So what do you do in your code is look at each pixel value, load all the pixels into array for this image, and you look at only the grayscale, color doesn’t matter at all. And then within these rectangles you take the sum of every pixel that’s in here. So you can’t see because it’s behind the black, but you take the sum of all these little pixels and average them together. Its the sum of everything in there. and then that is combined with all the other black areas in this and the black area is subtracted from the white area then you have your leftover value. These values are compared to thresholds in here. And then in here [indicating a node in the XML file] you can see a floating point number and in this case it’s -0.0315. So if you take the average of all the white minus the average of all the black you’re left with a value and if that’s with in -0.03 then good it passed that one and then it goes on.