A simple explanation (as I understand them):
Imagine the collision layers as a large two colour image (black/white) that has the same dimensions as the screen.
When you draw an image to the collision layer it converts all transparent areas to black and all non (or semi) transparent areas to white.
When you do a check for a collision, all bmax does is determine if any portion of the image you are checking overlaps any white area in the collision layer.
*edit*
I'm not sure this is 100% accurate, but no doubt someone will jump at the chance to correct me :)
Imagine the collision layers as a large two colour image (black/white) that has the same dimensions as the screen.
When you draw an image to the collision layer it converts all transparent areas to black and all non (or semi) transparent areas to white.
When you do a check for a collision, all bmax does is determine if any portion of the image you are checking overlaps any white area in the collision layer.
*edit*
I'm not sure this is 100% accurate, but no doubt someone will jump at the chance to correct me :)