Saturday, November 26th 2016

"Zoom and Enhance" to Become a Reality Thanks to Machine Learning

The one phrase from television that makes IT people and creative professionals cringe the most is "zoom and enhance" - the notion that you zoom into a digital image and, at the push of a button, it converts a pixellated image into something with details - which lets CSI catch the bad guys. Up until now, this has been laughably impossible. Images are made up of dots called pixels, and the more pixels you have, the more details you can have in your image (resolution). Zooming into images eventually shows you a colorful checkerboard that's proud of its identity.

Google is tapping into machine-learning, in an attempt to change this. The company has reportedly come up with a machine-learning technique that attempts to reconstruct details in low-resolution images. Google is calling this RAISR (rapid and accurate image super-resolution). The technology works with the software learning "edges" of a picture (portions of the image with drastic changes in color and brightness gradients), and attempts to reconstruct them. What makes this different from conventional super-sampling methods is its machine-learning component. A low-resolution image is studied by the machine to invent an upscaling method most effective for the image, in-situ. While its application in law-enforcement is tricky, and will only become a reality when a reasonably high court of law sets a spectacular precedent; this technology could have commercial applications in up-scaling low-resolution movies to new formats such as 4K Ultra HD, and perhaps even 8K.
Source: Google
Add your own comment

28 Comments on "Zoom and Enhance" to Become a Reality Thanks to Machine Learning

#1
Steevo
Meh, what happens when the machine is only as good as humans as juxtaposing what might be there, will we still place faith in what a machine with AI does when all it knows is averages, and puts an average face on a pixelated "cleaned up" image and gets someone innocent thrown in jail?
Posted on Reply
#2
slozomby
Steevo said:
Meh, what happens when the machine is only as good as humans as juxtaposing what might be there, will we still place faith in what a machine with AI does when all it knows is averages, and puts an average face on a pixelated "cleaned up" image and gets someone innocent thrown in jail?
i'd trust a cleaned up image far more than i trust the average human to remember what happened at a crime scene. if we're convicting people based on a single image then we're already screwed.
Posted on Reply
#4
R-T-B
slozomby said:
i'd trust a cleaned up image far more than i trust the average human to remember what happened at a crime scene. if we're convicting people based on a single image then we're already screwed.
Except we're not talking "cleaned up." We're talking outright creation of new data from adjacent pixels to fill in what should rightfully be nothingness.

Given that, frankly, I wouldn't trust it at all.
Posted on Reply
#5
LAN_deRf_HA
There's already ways to do this now. The latest "waifu2x" has a photo mode and the output looks just like that. Results are a real hit or miss but it works surprisingly well sometimes.
Posted on Reply
#6
RejZoR
If you've been living under a rock, this is something "new". If you know this tech from years ago, it's called S-Spline filter, developed by Benvista and used in their PhotoZoom.

http://www.benvista.com/photozoompro/examples

I've also used their S-Spline filter to optimize System Shock 2 cutscenes...

https://rejzor.wordpress.com/2015/05/26/system-shock-2-high-definition-cutscenes/

Clearly, you can't pull non existent information out of your ass, but advanced prediction algorithms can certainly elevate the image details far beyond high quality filters like Bicubic or Lanczos.
Posted on Reply
#8
natr0n
I bet emulation forum users are shitting the floor with this news.
Posted on Reply
#9
newconroer
Steevo said:
Meh, what happens when the machine is only as good as humans as juxtaposing what might be there, will we still place faith in what a machine with AI does when all it knows is averages, and puts an average face on a pixelated "cleaned up" image and gets someone innocent thrown in jail?
I believe that's why he mentions that a spectacular precedent would have to set.
Posted on Reply
#11
RejZoR
If I had the tools to quickly process textures, I'd use PhotoZoom on old games and scale them up to higher resolutions (if game engines support higher res).
Posted on Reply
#13
Prima.Vera
Vario said:

I always got cracked up of that scene, lol. The resolution of that film must have been, what, 512K? in order to pull that kind of detail.
Posted on Reply
#14
marios15
CAT-THE-FIFTH said:


:D
OMG hahahahahahaha this is gold and true in some years(?)
Posted on Reply
#18
jigar2speed
btarunr said:
I did not forget because it does not figure in the source article.
Agreed, your source link doesn't mention AMD at all...
Posted on Reply
#19
$ReaPeR$
its so nice to live in an age with such technological achievements!
Posted on Reply
#21
arterius2
BiggieShady said:
Oh yes, I have (re)played Day Of The Tentacle and other SCUMM games upscaled in ScummVM emulator using similar filters ...

AdvMAME3X looks pretty good.
Posted on Reply
#23
geon2k2
CAT-THE-FIFTH said:


:D
nice one :)
Posted on Reply
#24
arterius2
CAT-THE-FIFTH said:


:D
This is so sad, filmmakers, of all people should understand this considering they work with visual media.
Posted on Reply
#25
WithoutWeakness
The "watercooling-powered backplate" appears to be an acrylic reservoir. There are a few threaded inserts for tubing fittings on the backplate. You can see red and black wires attached to the backplate in the third photo which I would guess are to power LEDs. The front of the card is the waterblock plus aluminum fins and fans. It is likely very similar to what Asus did on their Poseidon cards. The waterblock on the front has a few threaded connections on the end piece that you can see in the third photo. This is where you would attach barbs at tubing for the GPU. There is likely a pump buried somewhere under that monstrosity of a 3-slot cooler or inside the silver block on the end of the cooler.
Posted on Reply
Add your own comment