You must have experienced images overflowing, hidden or simply overriding content on a web page. Dr Shai Avidan and Dr. Ariel Shamir get together to give us Content-Aware Image Resizing (CAIR) (pdf, alternate locations at Corel Cache, ACM). You can also see a video demonstration to experience the technology.
The idea is to remove the least significant bits of information when the image is down-scaled and add when it is up-scaled, instead of eliminating a random set of pixels. This preserves the relevant and important content in the image, giving us the best resized image. The key here will be the algorithm used to determine what to remove and what to add. Easier said than done, a Slashdot commenter does a nice job of outlining the possible algorithm.
It is an exciting development, but we need to give more time to it, to identify edge cases, raise questions and answer those. My first doubt is about the behavior when the less important information, that gets removed, is necessary and relevant. Having said that, the work is invaluable and I think can be applied in other image-intensive applications. Another big benefit will be for web designers developing liquid designs.
August 28th, 2007 at 5:27 pm
Interesting design. Anyway, I would suggest that you embed the video about content aware resizing on this page. That video makes it very easy to understand the concept.