Resizing images on screen has always been a headache -- inevitably data will be lost either through cropping or scaling the image smaller, or look significantly worse by having to generate extra data to get it larger.
A new technique for image resizing has been developed at the Efi Arazi School of Computer Science and recently displayed at Siggraph, the computer graphics trade show in San Diego.
Called Seam Carving, rather than cropping a bunch of pixels out in a big square chunk, the algorithm cleverly finds "streams", a path that can meander horizontally and vertically in any direction, connecting "low energy" parts of the image -- that is, unexciting parts with little change in contrast or detail -- and then strips them out, allowing non-uniform scaling to take place.
If things warp a little too much, sections can be easily marked to keep areas safe from resizing, or conversely you can raise the removal priority of an area for an easy way of eliminating that ex-lover from all of your photographs.
If you're saying "huh?" at the explanation (and frankly, who wouldn't!), the video here will help clear things up, and show how amazing this really is.
Potentially this technology could also be adapted to Web pages, finally ridding us of fixed image sizes that bork layouts, dynamically readjusting the field of view so the important information and image quality remains, regardless of screen resolution.