Few weeks back, I had written about Photoswapping images instead of Photoshopping them. Here's another groundbreaking algorithm for "content aware image resizing", thanks to Michael from Techcrunch:
[youtube=http://www.youtube.com/watch?v=qadw0BRKeMk]
The research paper, as well as a downloadable video, by Dr. Ariel Shamir and Dr. Shai Avidan is available here.
Note the differences between the earlier algorithm and this one. The earlier one used a large database of images to add or replace image sections. This one doesn't work with any external images, and is purely an algorithmic advance towards higher content awareness of the existing image itself. Now you can easily use photos from your Ibiza vacation to print family-friendly photo albums!
It is fascinating to see how we're making advances in algorithms even today - why wasn't this invented all these years that Photoshop has been around? (Simple answer: because no human mind had conceived and invented such an algorithm before).
I do wish the inventors get richly rewarded for this magnificent piece of work!
Technorati Tags: digital photography, image manipulation, image resizing, photoshop, image editing
| Share this post : |  |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
Related posts:
This is brilliant!
Er, what is it, actually?
Wow! This is indeed brilliant. Thanks for highlighting it!
Arun: thanks!
Rambodoc: I presume you’re joking…:-)
This is awesome. It would hours to do some of the stuff he is doing on Photoshop
No, seriously, this looks great and all, but what would I use it for?
Its not as fast as the videos show it to be. There is a lot of precomputation done beforehand(the energy map for example). The idea is cool but the performance on images with a lot of edges(high energy) needs to be seen.
doc — the guy mentions it at the beginning. Basically to allow for “more effective representation of the entire scene” in a smaller version which could be used for smaller displays like say on an iPhone
, and also allow for general web content (to accomodate slower connections better)
Rambodoc: Regarding what is it: have you ever tried resizing an image in any image editing tool? If you resize it in a diagonal direction, you can keep the aspect ratio intact and there’s no problem. But if you stretch or compress an image in only one direction — horizontally or vertically — it gets skewed and distorted.
This algorithm is the first one to let you stretch or compress an image in one direction without distorting it. It ‘intelligently’ fills in the empty space created while stretching, and merges the borders ‘intelligently’ while compressing. That’s how you can turn the photograph of a couple into that of a single person or expand the sky in a photograph for a better composition.
Regarding what you would use it for, I don’t know. I’m sure image artists will use it for a lot of the behind-the-scenes image editing that we are not even aware of when we browse the web, read magazines with cool photographs, and buy posters or greeting cards with beautiful-looking images. Your question is a tricky one: it was asked by mankind when someone first invented the wheel. It was what made the Product Manager for the IBM PC say that “there is a world market for maybe five PCs”.
Ah, now the post makes some sense. After all, the last refuge of the educated moron used to be the Medical school, so you know why I ask these questions!
Ah! How blessed I am to be in such enlightened moronic company!
That looks really cool. I can see how this can be helpful, but wow, did that go way over my head. way, way over my head. But, thanks for sharing.