Adobe's Jaw-Dropping Content-Aware Fill
Adobe's new eye-popping Content-Aware Fill feature looks set to make designers' lives a lot easier (and no doubt shift a few units of Photoshop along the way). Matt takes a look at what this new feature might mean for the design industry.
Just as Adobe gears up to launch version 5 of their Creative Suite, they drop this little bombshell on us — Content-Aware Fill:
Watch the video above to see what Content-Aware Fill can do. Removing the bits of trash and the tree is impressive enough, but wait till you see the road being removed at 3:53, and the extra sky and clouds appearing at 4:44! To paraphrase Arthur C Clarke, this technology is so advanced it's almost like magic.
Okay, you've gotten over the initial "wow" factor and managed to close your mouth again. How does this tool work exactly?
Content-Aware Fill is new Photoshop feature that resembles the Patch tool on steroids. It uses a very smart texture synthesis algorithm to sample image data from around the the area to patch, then reconstruct a texture based on the sampled data in order to patch the area.
Will it be in CS5?
Possibly. John Nack from Adobe cryptically talks about it appearing in a future version of Photoshop, which might mean CS5, or it might not. On the other hand, this video appears to demonstrate Content-Aware Fill - AKA "PatchMatch" — in an early version of CS5 (codename: "White Rabbit"):
I know one thing for certain though — if this is in CS5 I'll be upgrading for sure. :)
Will it live up to the hype?
The demo video mainly shows Content-Aware Fill working on landscapes, which tend to have large depth of field and fairly predictable, uniform textures. It'll be interesting to see if it can work equally well with scenes containing lots of contrast and detail, such as removing these urns from this shot of the Palace of Fine Arts in San Francisco:
That said, even if Content-Aware Fill has its limits, it will still be a massive help when it comes to retouching photos.
What effect will this have on the design industry?
Some might say that this feature will take away a lot of the skill of Photoshopping. I disagree. It will end the tedious process of painstakingly removing unwanted objects from a photo, which frees the designer or photographer to concentrate more on the creative side of using Photoshop. This can only be a good thing in the long run.
On the negative side, Content-Aware Fill will make it much easier to remove watermarks from stock photo comps, which may have an impact on stock libraries' sales figures. (Hopefully not.)
For me, the irony with Content-Aware Fill is that you'll finally be able to use Photoshop to remove lens flares. ;-)
Follow Elated
Related articles
Responses to this article
19 responses (oldest first):
I'm sure a lot of Photoshop users would be more than happy to have this feature. But as you've mentioned, let's hope it would also work well on pictures that have a lot of contrast and users don't use it negatively to remove watermarks from other peoples photos.
- mauco
If it works, a massive timesaver.
Simon
Apparently a plugin for GIMP already exists that does the same thing. Haven't tried it yet but I'll give it a go!
http://www.youtube.com/watch?v=hLtYPmRMYRE
Well, all you have to do is download Gimp and the plugin from the link I posted and you'll see it for yourself! The point for the same demo was probably to show that Gimp can do now and for free what PS promises to do in the future for megabucks? Just a guess!
You're right about one thing though... it is "magical"! I just tested it with Gimp 2.6 on Mint Linux and it works great! You just need to use Synaptic Package Manager and search for Resynthesizer to install (for those like me who are not used to the Linux way of things!
). to ones used elsewhere that are attributed to Gimp.
I grabbed the images from Adobe's vid and tried it out (Filters > Enhance > Smart Remove Selection). Here are the results:
Tree:
Original: http://img185.imageshack.us/i/treek.png/
Result 1: http://img214.imageshack.us/i/treeresynthtry1.png/
Result 2: http://img337.imageshack.us/i/treeresynth.png/
Road:
Original: http://img97.imageshack.us/i/road.png/
Result: http://img222.imageshack.us/i/roadresynth.png/
Panorama:
Original: http://img27.imageshack.us/i/panoa.png/
Result: http://img153.imageshack.us/i/panoresynth.png/
As you can see, the tree removal failed the first time, and included bits of the background trees. When I selected a smaller area that missed out some bits of the original tree, it worked much better.
The road worked really well - just as good as Adobe's version in fact.
The panorama wasn't as good as Adobe's though - mainly because it missed the ground!
But anyway - amazing stuff.
This takes the shine off Adobe's announcement somewhat!The GIMP developers should shout about this plugin more - I'd certainly never heard of it, and it's very impressive indeed.
@najobskalf: I'd imagine the creator of the Resynthesizer YouTube vid simply grabbed the images from Adobe's demo vid, as I did.
Cheers,
Matt
We need some coding wizard to step in and take the plug-in further. As I understand it from the Resynthesizer website, it is not being actively developed. Perhaps someone could improve it. Having said that, well done to Dr Paul Harrison for developing this for his PhD Thesis!
The possibilities are limitless. It would seem that the Adobe version is more advanced in development and more powerful
than the Gimp version. Just look at the following mindblowing demos:
http://videosift.com/video/Seam-Carving-for-Content-Aware-Image-Resizing-wow
http://videosift.com/video/More-on-Photoshop-Content-Aware-fill
The second one is most impressive. How can I have ever doubted that this is the real thing!
@najobskalf: Ha ha, that second one is brilliant!
See for yourself:
http://ilikekillnerds.com/NoVases.jpg
Simon
You'd have to use the spot healing brush set to a really small brush size and set to content aware to properly remove elements like that from images.
Post a response
Want to add a comment, or ask a question about this article? Post a response.
To post responses you need to be a member. Not a member yet? Signing up is free, easy and only takes a minute. Sign up now.
