Just Posted: Lytro Light Field Camera review and video

yes you can do it in everyday mode at so close a distance that the referee may halt the game to get you out.

the problem with Lytro camera is, while they tell you that you can refocus, you have to give out 90% of resolution and 9.5dB of SNR (more than 3 stops) first. what the hell a brilliant idea!

*: the lose of SNR can be partly regained later via software.

yabokkie:

Where are you getting these signal to noise numbers?

if you have any knowledge to share then please say it.
if you don't, then just read my previous posts and some by others.

yabokkie:

Well, I read all but the last 10 pages of Ng’s Phd and there’s nothing inherently noisy about a light field camera.

Nor is the image somehow much much “darker” as someone claimed here.

I happen to think Lytro made a mistake shipping this camera with the fixed ISO of 800, that setting could account for some “noise” and banding in Lytro examples. But almost any sensor of these dimensions with 16MP would have noise at that ISO setting. (And anyhow, as you hinted at, that ISO problem can be fixed with a firmware update by Lytro.)

Consider that added.

You still have not answered the question of what makes you think this system is so particularly noisy and where you get the specific signal to noise ratio numbers.

yabokkie:

It aint a 3D image, you really aren't showing any real familiarity with this technology.

You're still confused about signal to noise too.

yes I agree it could be worse than I thought. actually the reason why they decided not to go 3D could be that the effect is too weak.

search "Canon 120MP" APS-H sensor for surveillance.
currently the cameras are more resolution challenged.

each image a Lytro camera captures is 9 times darker (so it's really a "light demanding camera"). by simply stopping down your current camera 3 stops, you can gain a lot of DoF if that's what you want.

yabokkie:

No Lytro images are NOT nine times darker.

This Lytro camera is in fact pretty good in low light.

No, stopping down a lens does not give you the same result.

you got a deeper field at a much lower resolution, what's the point? anyone can instantly get 3 times "deeper field" by downsize the image to 1/3 resolution. you don't gain anything, only lose the resolution.

yabokkie:

You gain being able to fix focus after the fact, blur too.

yabokkie, imagine being able to go back to out-of-focus images and refocus them onto what you want. Use a smaller aperture? Well, what if I don't want everything in the scene to be in focus? That IS one of the reasons I shoot a camera with a large sensor, to have a narrower depth-of-field.

HowaboutRAW writes: "You gain being able to fix focus after the fact, blur too."

...because we really need your "cheesey - after the fact Gaussian blur" that uses a marquee'd area in the photo that's arbitrarily defined by your rudimentary contrast-sensing algorithm. Please, spare me now. I simply don't think photographers will fall for this feature that only YOU so highly covet. They'll want REAL optical DOF like the type shown in this picture from Canon's new 5D mkIII. Do your people simply not understand this simple fact of real optical blur?

http://web.canon.jp/imaging/eosd/samples/eos5dmk3/downloads/09.jpg

realityChecker:

You still haven't bothered to read up on Lytro tech.

Is uses real optical focus and real optical blur. Ironic that the the principle of the Lytro camera is that it processes more information than is readily available to a normal camera's sensor or film plane. (To clarify, that's what you're refusing to do--process.)

Erik Magnuson:

3.5x of what?

The sample that I played with did not spend any time focusing.

I have never seen a P&S with this zero shutter lag, even with autofocus turned off. And I've tried a good number.

Imagine a Nikon D3s, with autofocus off.

A good number of poorly executed digital cameras were released between 25 and 12 years ago. Look what happened.

High-compacts are not novelties.

Erik Magnuson:

And what's wrong with 43mm to 150mm equivalent? Many cameras have some limitation in a special mode, and this mode is not real desirable. In other words who cares, particularly on a first release?

This Lytro is faster to shoot than 12 year old P&S cameras.

No, Sony and Apply both withdrew digital still cameras from the market--20 to 25 years ago. Yep it's true.

> And what's wrong with 43mm to 150mm equivalent?

You need to convince the consumer that what he's getting for his money is worth it. 3x zoom and limited resolution for $400 was ten years ago. It also shows the optical limitations of LFC at the current level of decimation.

> This Lytro is faster to shoot than 12 year old P&S cameras.

If you take AF out of the process, then both Lytro and a compact have to do the exact same task: measure light, reset the sensor, and trip the shutter. Except for any difference in 12 years of electronic progress, that takes the same amount of time.

Are you suggesting Lytro will need to withdraw from the market at least once before LFC will become mainstream? (BTW, 25 years ago cameras were electronic but not digital aka still video cameras.)

Erik-

Quote:

"Are you suggesting Lytro will need to withdraw from the market at least once before LFC will become mainstream? (BTW, 25 years ago cameras were electronic but not digital aka still video cameras.)"

Yes, and Lytro may not be in the game in 25 years--too.

No, Sony really did ship a digi still cam in about 1987--not video.

Sony DSC in 1987? Which model? All of the contemporary Mavicas will still video cameras.

I'm not saying that the Foveon technology doesn't work, just that it has limitations that keep it out of the main stream. I can see the Lytro unit becoming another niche device - assuming they can bump up the resolution.

Is this for the Nokia 808 news? Hehe. They got rid of the mechanical things on that phone. It is fixed lens and fixed aperture. Focus can be set to hyperfocal or infinity.

I feel the future of this innovation is positive. Still photography is just one of the many areas that we can apply this innovation. Movie making, surveillance and distance measurement, court evidence, replacing 3ccds, marketing tools, presentations.... and many others.
But new technology is short-life, if they don't speed-up their process in making returns for their investments, I'm doubt how long the people still feel interest in this.

Michael Long:

Got a page number in the thesis for that 112MB sensor claim?

You have to give them credit for trying. It is exactly the concept that looks good on paper but fails as a product, but their optimism is commendable.
I think as a standalone device it is useless, but probably it could be pretty good for mobile phones as a way to cut the cost as an integrated solution.

It's a big difference between some dusty academics pondering their "old idea" in "schools" and someone that actually steps out into the blazing sun outside and make a product.

Now it becomes a question if Lytro is a Xerox or an Apple, i.e. if they are going to be the inventor of clunkiness or the company that turns the tech into something usable.

This not a 3D camera.

> I'm still wondering what the "resolution" might be

yes Lytro has a foundamental problem in marketing. the optics and components are basically all the same and there is nothing special but they can't say it or their product will taste flat and look poor.

Michael_13,
it's effectively a 3D camera made to output a series of 2D images only. it should not be difficult to produce a 3D image on your TV. straight forward thing.

JadedGamer:

Xerox tech is the basis of Macs, the iPhone, and Window.

Irony that you picked that example.

That Xerox was stupid and gave away the GUI is a different story.

I think that is exactly what it could do with suitable software. You could shift the planes of focus just as with a view camera. Or even have two intersecting planes in focus and nothing else. (Within the range of focus of the lens as the microlens/lightfield design provides the equivalent to several extra stops of depth of field from a fast lens.)

Future developments of this technology and software should make it more useful as a creative tool. But this camera is certainly not doing much with it yet. Targeting the p&s market seems kind of pointless considering what cell phones and cheap cameras can already do.

@infosky, think of it like this: apply focus stacking techniques to the series of images produced for different planes. For example, you could combine the closest planes with the farthest planes yet leave the middle ground OOF.

You could get some depth information from multiple AF sensors where the focus aperture is different from the taking aperture and/or the changes during the focus operation. This may be enough to do a credible job for most artistic photographic purposes.

Samsung just announced a new sensor prototype that records z (depth) information along with the traditional RGB mosaic. Once developed further, that would give you the ability to emulate wide-aperture background blurring via processing.

It's interesting to hear that selective defocusing technologies are indeed under development. To Stephen123's point, yes what I'm talking about is in essence standard Photoshop--selecting regions of the image and blurring them. What I'm wondering, though, is if there isn't a way to more or less automatically accomplish this via a firmware option rather than manually, after the fact, via software. When you depress the shutter halfway to focus, on, say, a person in the foreground, the camera then knows to blur out every other aspect of the picture to the degree the photographer desires. To me, this approach seems inherently easier--and doesn't require the image quality sacrifices of the Lytro--than the very elaborate microlens array. To apply this effect in a nuanced way I suppose you'd need what Erik suggests, multiple AF sensor readings, but that still seems easier than the microlens approach and wouldn't compromise the quality of the actual image capture.

But what do I know, really.

And, they were right at the time. Digital wasn't anywhere near ready for prime time...hardly even for consumers. In time though, it took over.

We'll see how this technology does 10 years from now.

> Bracket focusing can't really do say 10 foot depth of field.

Yes it can - it's called focus stacking. http://en.wikipedia.org/wiki/Focus_stacking

No, I get it. You said "can't"; it can. Doing it quickly would require a lens and focusing algorithm designed for that (e.g. small internal focus element that can move quickly to different steps.)

Much of what Lytro does can be done with multiple exposures -- just with different implementation and computational considerations. (Hint: what do the prisms/microlenses effectively do except record multiple different focus views of the scene? The difference is in the simultaneous capture and the reconstruction processing.)

Erik Magnuson

No, much of what Lytro does can't be done with multiple exposures, for example moving things.

I should have written that focus bracketing only emulates Lytro in very very limited circumstances.

"Moving things" are not "much of what Lytro does" if you look at their sample photos. Even then, it depends on the composition: for example the Lytro sample of the surfer eyeing the waves would only require 2 or 3 focus planes to emulate.

"Lytro" is a stand in for the possibilities of light field photography.

Really, way to be purposely obtuse.

Resolution test shots should be done though. I doubt the 1080^2 resolution, I guess its is upscaled. The preliminary math I did tells me so. It yields more like 820^2 assuming 4x4 rays per pixel which is also a plausible figure taking the Bayer filter into account.

Moreover, I notice that ISO 800 hapens to be 4x4x50.

falconeyes:

Not sure there is a Bayer filter.

Do you have the Lytro software installed on your Mac to view and manipulate the files? And do you happen to have Lytro files from somewhere else?

No, I only know what is written and the FCC teardown. But Lytro says it is a standard CMOS sensor with a special microlens array. I see no other way to record color then.

falconeyes:

I'm going by my memory of reading Ng's PHd some months back, but don't see the text at the top of my pile of books, so can't look quickly. I don't remember a bayer filer with the Contax he modified for the dissertation research.

Remember that prisms split light.

Other light field papers/implementations suggest that the output "resolution" actually depends on the focus point. On the dpreview sample with the 3 cameras, the Leica in the back has noticeably poor detail.

@HowaboutRAW : It's a special micro lens array placed higher than one normally would be, over a standard bayer sensor.

One thing to note is that he mentions if you could change the distance between the micro-array and the sensors, you could have a "normal" camera that's could also be switched into "Lytro" mode.

@falconeyes : It's ALL interpolated, but it might well be higher than you think. Part of the process corrects for lens distortions, chroma, and other aberrations, all at the same time.

I don`t think you need phd to uderstand this http://www.lytro.com/renng-thesis.pdf

The thesis was submitted in a computer science school. Why not physics or optics? Did a camera engineer review and comment on the document? Did it appear in a peer-reviewed journal open to comments by other camera designers?

No, not a PHD, just a PDF ;-)

Cy Cheze:

You can read the thesis and find out why.

It's more like the computer application of this technique, which Ng did not invent. There is no new optical science.

He made it all work with a medium format Contax and a laptop.

@Cy Cheze: You might go on Flckr and note just how many photos come from traditional camera makers vs computer and electronics companies...

@HowaboutRAW - You said "An obvious application, with much better video gear, would be to shoot a whole movie and not worry about focus while shooting--ever."....Really?...OK so clearly you (or your PR people) know nothing about film making because no DP would ever want that. Ever! First you guys need to learn about why we like our 7D's and 50D mkII's and why we've replaced our video cameras. Remember, find a need, fill a need. That's biz 101.

realityChecker:

You've proven that you have no idea what Lytro does.

And why on earth are you bringing up DSLRs that have been 20 years in development?

Of course a DP would want to be able to shoot and not worry about focus until processing time. It is you who has proven your ignorance of movies and tv and sports.

HowaboutRAW: I'm with robogobo (above) who says "No matter what research is pointed my way...<snip>"

Real photographers & film makers don't care about your prism technology. It's a gadget. Why am I bringing-up DSLR technology that has matured 20 years? Because it's the current market leader and the reason we fork-out thousands to the likes-of Canon and Nikon. You think we can't get a shot infocus at f22 and a high ISO setting on these things? It's you that are showing ignorance. Watch this recent video by Canon Europe on the new 5DmkIII and then tell me that "post processing" the blur actually makes ANY sense? http://vimeo.com/37804617

I mean please, do you really think that the DP wished that he could do the BOKEH AFTERWARDS on every frame that needed it with a bank of servers to chomp on those bits like a 3D rendering? That's ridiculous! Do you think he or she had any problems getting the focus right? You're clearly not a DP just sounding more like a hired PR person or an employee.

Doesn't really work for that but you could capture the movement by focussing different spots, which would be interesting. Disadvantages always have an upside ;)..

@HowaboutRAW: Moving object? Sure. No problem. You want to capture 10X the data PER FRAME in order to do so?

We barely have the tech to capture 4K. You want to try 40K???

Interesting thought about built-in "non"-replaceable batteries, which (of course) ARE replaceable, just not by users ...

... I have NEVER had to replace ANY of my Li-ion batteries in a mix of laptops, Nokia phones, and assorted digital cameras. Invariably, the DEVICES have failed or been replaced due to becoming obsolete, LONG before the battery expired. The reason is, I top-up-charge my Li-ions after EVERY use, and this way, they last for very many years.

The same was NOT true about NiCd, or even NiMh, but thankfully those days are passed.

So am I worried about built-in batteries in my iPad 1? No, because by the time they go, I'll own an iPad 5. And my lovely Olympus E-1 is still going strong, 10s of 1000s of photoes later, after 6+ years of use on the same, original battery pack.

Yes, I've just touched wood to avoid cursing my kit, but let's face it: you don't NEED to be able to replace Li-ion batteries in kit with an expected useful lifetime measured in just a handful of years.

I think it's down to how you use your Li-ions. They DON'T like being fully discharged, and should NOT be routinely run down and then charged - the exact OPPOSITE of NiCd and NiMh. If (like me) you keep your Li-ion devices topped up all the time, the batteries will last you for years.

Brian

Just checked - Apple provide a battery replacement service for all their iPods, iPads and iPhones, at reasonable prices. So after all, the built-in (or rather, non-USER-replaceable) battery is NOT an issue!

Felix E Klee:

The Lytro performs surprisingly well in low light.

Felix E Klee:

No most of the light information is kept, that's how they can do the refocus.

Note this camera ships set at ISO 800--that's high. I happen to think that Lytro probably overreached a bit.

Camera shake is not removed by this technology, shutter speed does that.

Ivanaker:

To clarify: The Lytro does NOT have a special sensor--it's off the self.

It's the prism and software that are particular with this Lytro. Nikon or Canon, or any lens maker, could work on a lens that would also do this. But that's going to be more expensive than the prism.

ok, its not the sensor, but anyway, i welcome new tecnology and cant wait to see what is going to grow out of this.

Felix E Klee:

If you are talking about printing a jpeg, yes then information is discarded.

But that's not how the Lytro files work. You need to think not of pixel by pixel resolution, but another dimension of resolution. Yes the 2D image isn't that many pixels. No sure that this version of lower resolution avoids some camera shake problems, because the resolution of this sensor is used elsewhere.

Felix E Klee:

Right, go to the Lytro website and click to refocus the images, that's not discarded information. That's information which is being rearranged. To print a jpeg is discarding information, as I already wrote.

Read again what I wrote.

Felix E Klee:

The information in the in-focus area is not discarded from the file--by your explanation, if true, the information is suppressed in the in-focus area of the photo representation on the screen.
Then a different in-focus are can be selected and different information suppressed.

So no the information is not discarded as with a jpeg.

Never said this. Read again what I wrote.

Felix E Klee:

You from above:

"Unless there is some magic going on, most of the light information reaching the sensor is discarded when rendering the picture."

Felix E Klee:

Quoting you from above:

"Don't see the point. As far as I understand it, when rendering a picture, the Lytro throws away most of the light rays reaching the sensor."

Please don't make the mistake of assuming I can't read.

Felix E Klee:

Quoting you from above:

"Don't see the point. As far as I understand it, when rendering a picture, the Lytro throws away most of the light rays reaching the sensor."

Please don't make the mistake of assuming I can't read.