Possible solves (from easiest to hardest): * if, walls or paths are a flat color, use that color in your collision detection; * if wall or path are favored with much more red, green or blue, you can use that in your color detection; * otherwise, add Wall actors or Path actors to be used with collision detection; With the first two ways, draw maze on background image, otherwise, there would be maths to deal with to determine which pixel to check (dealing with location and size of maze actor).