Microsoft algorithm uses six-axis motion sensors to fix blurry snapshots, inadvertently pimping your ride
Trying to snap a shot of your cherry red Mazda, but can't keep your hands still? You'll find all the tech you need to smooth things out in an iPhone 4 or (MotionPlus-equipped) Nintendo Wiimote. Experimenting with 6DOF inertial measurement sensor packages, scientists at Microsoft Research have developed a software algorithm that literally records your exposure-destroying shake via accelerometer and gyroscope, then magically removes the blur by canceling it out. While the technique still isn't perfect -- spot ghostly line above some of those background cars -- the Microsoft researchers compared their results to other in-progress algorithms, and we think you'll agree this new solution presents the best results by far. It's a shame Microsoft doesn't say when we'll see the tech in a spiffy DSLR attachment, or better yet a cameraphone. See before and after animated GIFs after the break, and find high-res comparison images and much more at our source link.





































Finally, we'll be able to get some decent pictures of Big Foot.
They should just embed the data in the exif and let Adobe do the work for them.
Microsoft:
Dear Sir,
Please fix my portrait.
http://is.gd/dYCtR
Yours, very sincerely and respectfully,
A. Lincoln
they just need to make better low light sensors for phones!
this is what they should have done in iphone 4. pretty obvious solution if you ask me.
Nice, want.
mmm, mr blurry cam is out of focus, not just blurry...
Those animated pics are tripping me out
I need this on my phone very badly. I would say at least 25-30% of the pictures I take get ruined by blur.
That is a 3D version of averaging. Cool. (I used to work in image processing.)
@Oflife
Actually it's the opposite - kind of 'un-averaging'. It's called deconvolution.
People still put bras on their cars?
Cool stuff - makes me wonder how current image stabilizers work in digital cameras.
I guess that will make its way into machine vision where motion-induced blurs are more predictable with machines than humans.
@GSGeek
I believe that in SLRs with the image stabilization in the body itself, the actual sensor moves in the opposite direction as the shake, which hopefully neutralizes any blurriness. For lenses that have VR, the same thing is applied.
Nice, now I'll be able to take clear pictures using my Samsung Vibrant of my Mazda RX8.
Mr. Blurrycam gives two thumbs down. He's an artist, technology just gets in his way.
Seems artificial....More than "fixing" a blurry shot I assume this takes the photo and assumes what it should look like when not blurry and outputs that image?
Thing is I can't imagine a system that takes a blurry photo and somehow creates details that were never captured in the first place. I suppose this would work for tiny 4x6 prints and your facebook profile but anything with a decent resolution I feel is gonna just look like it has harder lines...not more detail
If they figured out how to ADD detail after a photo is taken though...that would be something
@InnocentEd
A photo that is blurry because the lens was out of focus cannot be fixed -- the detail is lost (because the sensor never had it to begin with).
But if the picture was in proper focus, and the only problem is movement of the camera, then it seems reasonable that some algorithm can be applied to reverse the artifact. I'm sure it's imperfect, but it's not necessarily inventing any detail.
@greyseal I suppose that is true, but technically when you get a blur like that even if the lens is set to the proper focal length you are essentially un-focusing the shot by moving so that senor can't pick up the detail.
Anywho, I guess it's better than a blurry show :)
Dear RX-8 owner,
1985 called, and they want their front-end paint protection fad back.
Why didn't sony think of this! LOL
@Vincent180
http://www.sony.net/SonyInfo/technology/technology/theme/alpha_01.html
I actually think they did, quite a while ago. Just like any other camera manufacturer with a bit of self respect.
Some iPhone app claims that they use the accelerometer as a stabilization. Obviously a gyroscope would help more. Anybody knows if Apple has this tech in the iPhone 4 camera app?
@pika2000 I actually suspect Samsung use the Gyro in the Galaxy S for stabilization. The quality of night mode pictures on the Galaxy S are unusually good.
Can we make this a WP7 API now?
I really don't see why you can't record the incoming exposure over time, measure the movements, and then crop to the common center for the final image.
@maztec That's called digital stabilization. It "works", just not good, because it takes to much processing power I guess. Also, a computer is really not intelligent enough to see 100% how to join pictures properly. A gyroscope is a much smarter solution.
@HPe : Thanks for the info! However, I do not see why the two cannot be combined for greater accuracy...
The method detailed here is using the gyro information to correct blur after the fact. Whereas, digital stabilization + gyro would let the algorithm track center, and correct.
@maztec I think everyone want away from digital stabilization. It's the function used in mobiles and cheap digital cameras today. And it doens't work, simply because the software isn't intelligent enough to determine what's moving in the picture and what static.
@Engadget: Galaxy S can do it too , please add it to your article...
http://android.hdblog.it/2010/07/20/galaxy-s-sensore-di-movimento-a-6-assi-facciamo-chiarezza/
or
http://www.handheldusers.com/forum/t7232.html
I WANT THIS NA0. Google Googles is hopeless, and this feature on a phone would help! Is Google Goggles out on anything but Android yet btw?
I wonder if this will work for webcams too! That would be so cool. To break down the amount of noise while shaking. I guess that would make the picture all laggy then :-( nvm.
@affan of course it'll get laggy. The picture is blurry because the exposure time is extended, i.e. fewer frames possible per second.
Finally!
Apparently this technology has been around since 1998. Because no one has put a bra on a car since then right?