yeah i know there's some restrictions when using this method. I've read the code that does this. it basically receives a png file , x, y and alpha limit by parameter. It then stores all pixel data in a matrix, reads it pixel by pixel beginning on position x,y , left from right , top to bottom, tests alpha value of pixel , if greater than alpha it marks it as opaque. In the end it returns all rects corresponding to the sprites in the spritesheet. How exactly the algorithm works i couldn't understand
one implication of this algorithm is that all sprites must be alligned by Y. example:
<img src="http://dl.getdropbox.com/u/2472278/spsheet.png">
it would get the purple square first, not the blue one as it is expected, cause it is above the others