Here’s a mind-blowing new image-processing paradigm (yes, I hate that word too, but this really is a new paradigm). Researchers at the NYU Media Research Lab have built a trainable image filter. You give it an original image (say, a photo of a pear) and a “filtered” image (say, a watercolor painting made from that photo), and it analyzes the steps it needs to take to transform the original to the modified version. Thereafter, you can give it any image and it’ll “filter” it according to its derived rules. This is the ur-filter, the self-modifying code that learn from any example you present to it. Wow. Link Discuss