Works by cutting image into 1 pixel wide strips and discarding half. Repeats the cutting process both vertically and horizontally, reducing memory used by %50 each iteration.
Before being cut to strips
By adding the cut strips we keep the origional resolution of the image. Only after image compression we take a single copy of this compressed image and save it in its lower resolution.
