Hacker News new | ask | show | jobs
by alilja 4264 days ago
The fact that he's just using canny suggests to me that it's just a simple hack he threw together. There are LOTS of other edge detection algorithms, but canny is pretty good and so easy to implement that it's the most generally used. It's not great at faces at a low resolution, but even just a little pre-processing can make the image better (as he demonstrates towards the bottom of the page).
1 comments

Yes, 'simple hack' is definitely one reason. But there are a couple of others:

- The drawing is sent to the Arduino as a block, then drawn, so it has to fit in the Arduino's RAM. More complex drawings would require streaming the data during the drawing. That's totally possible, but it is more work.

- We quite liked the way these severely-simplified sketches came out. Because the edge-finding and vector-simplifying code is quite stupid, the best pictures tend to be the simplest ones, and doing more lines might well just make the picture more noisy.