PDA

View Full Version : more WOW...



omeone
08-30-2007, 12:01 PM
Can we have this in SE please? :D

http://blogs.guardian.co.uk/technology/2007/08/27/images_resized_just_how_youd_like.html

Really though, would love to see this becoming a mainstream image format.

art
08-30-2007, 12:26 PM
This is really amazing. I wonder if the algorithms would work for video too. They do not mention that in video (I did not read the paper)

omeone
08-30-2007, 01:02 PM
I would think so, since it is all digital, but would it be practical?

Didn't read the paper either, but I suppose the big question is how much processing it takes to make the image and then how much memory it uses.

Lightwolf
08-30-2007, 01:17 PM
This is really amazing. I wonder if the algorithms would work for video too. They do not mention that in video (I did not read the paper)
Hm, what it currently does is basically flag lines running through the image (not straight, but jagged lines, but always one pixel high/tall) with their importance.
Then these lines are removed/added (down-/upscale) depending on their importance.
For video the lines would need to be identical for the complete sequence.. both in importance and position/path - which I assume makes finding them a lot trickier.
If you had them re-evaluated for every single image, then you'd have a shift in your video as the importance of certain areas changes.

Apparently it is quite fast once the initial analysis is done, it just needs to store the paths and their importance.

Neat trick...

Cheers,
Mike

Titus
08-30-2007, 08:42 PM
They presented this paper at SIGGRAPH, very cool.

Red_Oddity
09-05-2007, 04:30 AM
And didn't Adobe hire this guy and his papers?
I rememebr reading a news article on a blog somewhere.

Found it : http://blogs.adobe.com/jnack/2007/08/imaging_heavy_h.html

So, either it disappears in the patent freezer or gets incorporated in a Adobe style barely usable interface-for-average-joe with a bloated install.

joseburgos
09-05-2007, 06:04 AM
This looks great.
Could possibly be used to auto removal subject when creating a 2D-3D photomontage.
Can't wait for it to be released to experiment.

omeone
09-05-2007, 06:55 AM
@red_Oddity

:( this is bad news