« OOXML overview | Main | Faster video compression »
September 02, 2007
Content-Aware Resizing implementations
Content-Aware Resizing implementations; Thanks to Mario Klingemann for tying together these links: Patrick Swieskowski, Ebert Joa, Lee Felarca and Hector Yee. Patrick has an online demo based in SWF; the rest do static image-processing in OS code. The set also includes links to other examples. Problems are mentioned with computational speed, algorithm fidelity, and interface conventions; the original SIGGRAPH reports mentioned unresolved problems with large image sizes. One of the ambiguities in this work is where the usage occurs: some implementations handle it like histogram adjustment or JPEG compression, at production time -- other implementations handle it at runtime, similar to resizing text in a browser. For full Flash work the effect is easy to fake, by separating out image planes at design-time... one layer a flat ocean, the beach on a layer above that, the rocks a third layer and people on the fourth... run the algorithms like an animation camera. This wouldn't recalculate the browsers' OBJECT/EMBED tags, though, so it's not enough to accommodate text reflow and window resizings. The best quality pixel-removal would be complete image analysis during production, but I think there's a need for realtime work too. There's a lot of ways to approach this problem, and it's great to see so many people working at it so quickly. Lots more breakthrough imaging going on these days, as computers learn more of what an image plane attempts to describe.
Posted by JohnDowdell at September 2, 2007 08:43 AM
Trackback Pings
TrackBack URL for this entry:
http://weblogs.macromedia.com/mtadmin/mt-tb.cgi/9031