This video presentation from Siggraph 2007 has been popping up all over the internets the last couple of days, and the implications are truly astonishing. Algorithmically this is a remarkably simple technique, and easily implemented in real-time. It should be pretty straight-forward to write an implementation in ActionScript 3 (for Flash 9) or in IronPython (for Silverlight) and have this apply to images in webpages with a minimum of effort.
More exciting than simply just resizing images is that the weighting and treatment that can be applied manually to specific regions of an image. It’s easy to imagine a myriad of gaming opportunities that arise if you can hide data selectively in images across the web through this technique.
Dr. Shamir’s other research work is pretty interesting as well, covering as he does:
- Mesh Partitioning
- Skeleton Based Representations
- Multi-Resolution Models
- Object Feature-Space Analysis
- Digital Typography
- Visual Succinct Representation of Information