Unbelievable Photoshop sneak

[dkcoats]

I gasped audibly.
Link

[Dropt]

This an incredible feature, but my problem with all that is, soon people will know we don't need to take good pictures to create a good one out of photoshop. And that worries me a little.

[John Wheeler]

IMHO

The unblur feature would be welcomed by me yet it is not promised to be in CS6. I would consider it another tool in the toolbox. Like all tools they have their limitations. A Master Carpenter can do great things with great tools (and also basic tools). Yet, even with great tools, I would not want an apprentice to build a house for me.

Think about the value of Content Aware Fill. That was touted as a super tool and for what it does, it is super. Yet, it does not solve all problems nor in some cases is part of the solution at all. The deblur feature will no doubt be like that. When you get an image from a client, this tool may be a productive way to help out. In other cases it may not.

As always, the best image always will come from getting the best shot to begin with. I am looking forward to giving it a test drive.

[plugsnpixels]

It has been revealed that Adobe "faked" at least part of this presentation, using an artificially blurred photo to "de-blur".

[chet webley]

That certainly is an impressive stunt.

I Agree with John, above - It will be another tool to help us achieve better quality images. I also thought that the final comment (by whoever that very irritating man with the red trainers on is) "Would I pay for it? No!" was very pertinent.

On a similar note, there was this rather interesting article on the BBC website this morning:-

http://www.bbc.co.uk/news/technology-15383516

[Chain]

The technology is absolutely useful, and I saw it discussed in some paper from 2006. Nice to see it getting better and Adobe starting to implement it. What it does is reduce blur due to camera movement (motion blur). How good the result is mainly depends on how accurately it is able to calculate that blur kernel (how the camera moved/the "path" of the blur) – so if Adobe in their example used artificial blur (with a perfectly known blur kernel) that would have given them a bit better results in the demo that what you can expect on "real" blur.

Chet: The light-field technology you link to is sweet (allows you to change focus in a photo after it was taken). It believe it generally uses an array of micro-lenses above the sensor. The downside is a lot lower resolution (and/or less light). I remember seeing a nice demo-video a while back.

[clicks2]

Originally Posted by Dropt

This an incredible feature, but my problem with all that is, soon people will know we don't need to take good pictures to create a good one out of photoshop. And that worries me a little.

I agree. But not just Photoshop, now this Lytro camera is coming out [looks like children's toy]. Crazy tech that makes an image near flawless.

Yikes

[plugsnpixels]

Chain, I wrote about this technology in a grad school paper back in 1992-! I saw an example somewhere that showed a blurred license plate being pulled back into focus. I can post quotes from my paper if anyone's interested.

[lkroll]

I saw the refocusing demo video a week or so too. It looks beyond simple deconvolution schemes. Not totally convinced that what they are achieving would be real-world compatible since the did use an artificial blur meaning they could have fudged the result (call me a sceptic). Still, if this technology proves real, why use a lightfield camera when refocusing can all be done in software.

[Repairman]

I love all this stuff BUT, as John W alludes to, you need to have an 'eye' in the first place to judge whether the results are convincing or not. As a commercial retoucher I haven't had many requests to 'unblur' pictures but no doubt a market exists. Hope this isn't going to make some photographers even lazier than they appear to be right now!

[2mittsphotog]

Amazing Technology! For that rare shot that was perfect but the shutter speed was a tad slow (thinking live event shooting) who knows, it may come to the rescue. We shall see. Great post btw!

[RobertAsh]

Originally Posted by Chain

Chet: The light-field technology you link to is sweet (allows you to change focus in a photo after it was taken). It believe it generally uses an array of micro-lenses above the sensor. The downside is a lot lower resolution (and/or less light). I remember seeing a nice demo-video a while back.

Actually it's much better than that. The technology behind Lytro is ray-tracing. It's similar to how ray-tracing in 3D animation hugely, hugely simplifies high-resolution modeling by eliminating the need to exactly model the surface. Instead, the software traces the paths of light rays from the viewer back to the object then back to the source.

Lytro uses the same principles to figure out at any given focus point how to trace the rays of light back to the source in such a way that will keep Object X in the photo in focus, then recalculates everything when the focus point changes to Object Y. That's how it literally allows you to determine focus after the image is captured. Very cool stuff.

The only reason it's low resolution is because this is Version 1. Just think back to when digital cameras we currently use had VGA resolution in their early day. They don't have VGA resolution any more, and neither will Lytro in future releases.

[Chain]

Originally Posted by RobertAsh

The only reason it's low resolution is because this is Version 1. Just think back to when digital cameras we currently use had VGA resolution in their early day. They don't have VGA resolution any more, and neither will Lytro in future releases.

The reason it is low resolution is because it needs several pixels on the sensor per pixel in the final image. (In a demo I saw of such a camera four pixels were used). This effectively will make it always have an image resolution lower than a regular camera with the same size/resolution sensor unless/until some new revolutionary sensor design is invented.
However, modern sensors can be very high resolution, so what remains is the question of how much resolution you need for the task at hand (assuming they iron out the other image quality/optical challenges).

[lkroll]

I really believe that in the future, they will be able to put over 100 sensors of the same resolution as today's best sensor into the same area as one sensor is today. Parallel sensors will lead to all kinds of abilities only dreamed of now. The only drawback is that it will require astronomical disk space to store images compared to today unless a new lossless compression technique can be created, but HDs are getting bigger and cheaper. Heard the other day that they figured out a way be able to actually create an 18 Terabyte drive; that's outrageous. Also, the complexity will require a lot of CPU/GPU power to process so that is something else that will be developed too. Still, most of the time, a simple jpeg is good enough for me. lol

[chet webley]

Originally Posted by lkroll

Heard the other day that they figured out a way be able to actually create an 18 Terabyte drive; that's outrageous.

Crikey! The computer I first used for creating computer graphics on had a 40Mb HDD!!

[Stephenoit]

Originally Posted by Dropt

This an incredible feature, but my problem with all that is, soon people will know we don't need to take good pictures to create a good one out of photoshop. And that worries me a little.

We are all ready at that stage. The most important thing is the software. Anyone can take a simple point and shoot camera and take good pictures