Termination: For each iteration we record the number of unsuccessful stroke placements (no stroke with sufficiently high quality was found within 1000 random positions). If too many failures occur in a sequence we stop the iteration (see Algorithm 1, line 1). This kind of stopping criterion seems to be much more stable than other methods such as root mean squared pixel errors (RMS) between target image and canvas. This is due to the fact that a lot of visual noise is created during our approximation leading to unstable image-based termination criteria.