You can blur already, can’t you?
If you were to increase the distance from the shape to the blur, though, you’d be left with the ability to draw with a single, blurry brush:
Once you have a big, blurry brush to draw with, you can draw the pixels of the image you’re trying to blur, and voila!
A a blurry image! As far as I can tell, this is the first time that this approach has been used to construct blurred images. It’s nice to be first!
Except it’s actually pretty slow. A 20-pixel blur on a 640×480 image takes my computer about 15 seconds to render. It’s not totally useless, though – it’s good at making a big blur of a small image that needs to be magnified, and it opens up the possibilities for a selective blur, like in a depth-of-field image. If nothing else, it’s an entertaining exercise illustrating that you can have a lot of fun exploiting the drawing API for actions it’s not really intended for. Until webGL is available for most people on the internet, we’ll need to resort to hacks like this. I can live with that!