Edge based image division for greater accuracy Could be a mix of current method with something that increases accuracy