The below video demonstrates a method of resizing images that leaves the important areas intact. The software analyses an image for the least important curved lines of pixels, based on a gradient magnitude¹, then adds or removes them as necessary. This results in apparently seamless (at youtube resolution, anyway) real-time contraction or expansion of the image area. Easier to see than to explain:

I’m sure they’re using images that suit the process, but it’s impressive nevertheless. They’re surely talking to Adobe about licensing…Found via Waxy.

1. I’m not entirely certain what this is, but I’d guess it’s related to the average brightness/colour difference between neighbouring pixels? [↩]