The point of Lytro

I looked at the images and thought "Brilliant stuff!"

But, after playing with the photos for 10 minutes, I got bored.

Can't see it being much more than a novelty.

(But then, I don't like Facebook, either).

--
markeast
 
The pictures can be transferred to Facebook from Lytro.com, but not directly, and cannot be modified except by Lytro's own software - ie, not within Facebook (unles Lytro later licenses the software to Facebook).
This kind of limitations are unfortunately true for all propriety formats, e.g. RAW formats.

--
Roland

support http://www.openraw.org/
(Sleeping - so the need to support it is even higher)

X3F tools : http://www.proxel.se/x3f.html
 
If I'm right, we have a 540x540 pixel image, albeit with adjustable focusing.

That would explain why there isn't more tech spec on it, why it isn't trying to look like a camera, and the 'order before you see it' sales technique...

This will be a fun gadget, and some will be happy to spend four or five hunded dollars to show it off, I think I'll pass until one comes out with a few Mpixels...
I guess you will lose a factor of 20-100 x - so a 100 MP sensor will give you a 1-5 MP image. Is that OK?

So - as long as they use a small sensor you will get way below 1 MP images. You may expect VGA resolution. Is that OK?

I think it works like this. Lets assume we have 5x5 sensor pixels per Lytro pixel - then you lose a factor of 25 in pixel count. This also means that you can play with the DOF as you have taken 25 images with F10 with your F2 lens, assuming your lens is square. But its round so you lose some.

--
Roland

support http://www.openraw.org/
(Sleeping - so the need to support it is even higher)

X3F tools : http://www.proxel.se/x3f.html
 
Anybody think that isn't possible?
Its possible - just imagine a high speed film camera. But the speed has to be very high to avoid problems with moving objects.
Anybody think that is what the Lytro is doing? (I've no idea)
It isnt. Its a plenoptic camera.

A plenoptoc camera is much easier tro make than your propsal.
Darn, nobody pinch my idea please, just forget what you have just read or I will be obliged to kill you...
Actually - your idea (which is not new BTW :) ) has a big advantage. You can make a 5 MP camera easy that way.

--
Roland

support http://www.openraw.org/
(Sleeping - so the need to support it is even higher)

X3F tools : http://www.proxel.se/x3f.html
 
Anybody think that isn't possible?
Its possible - just imagine a high speed film camera. But the speed has to be very high to avoid problems with moving objects.
Anybody think that is what the Lytro is doing? (I've no idea)
It isnt. Its a plenoptic camera.

A plenoptoc camera is much easier tro make than your propsal.
But apparently limited to such a small resolution they don't want to talk about it :-(
Darn, nobody pinch my idea please, just forget what you have just read or I will be obliged to kill you...
Actually - your idea (which is not new BTW :) ) has a big advantage. You can make a 5 MP camera easy that way.
I do wonder why focus bracketing isn't implemented on some of the DSLRs (at least not to my knowledge. It could be so useful when shooting at f/1.2 or for macro work.

Canon introduced in the Powerwhot Pro1, a camera that had some extra goodies in (focus bracketing, time lapse...) that are starting to reappear in some bridges and could so easily be added to DSLRs, it would just have to be adapted to cope with different lenses.
 
If I'm right, we have a 540x540 pixel image, albeit with adjustable focusing.

That would explain why there isn't more tech spec on it, why it isn't trying to look like a camera, and the 'order before you see it' sales technique...

This will be a fun gadget, and some will be happy to spend four or five hunded dollars to show it off, I think I'll pass until one comes out with a few Mpixels...
I guess you will lose a factor of 20-100 x - so a 100 MP sensor will give you a 1-5 MP image. Is that OK?
A couple of MP would allow full HD displaying and small to medium prints, that would be a big step forward.
So - as long as they use a small sensor you will get way below 1 MP images. You may expect VGA resolution. Is that OK?
It's fine for web page inserts, which is what it is being marketed for ;-)
I think it works like this. Lets assume we have 5x5 sensor pixels per Lytro pixel - then you lose a factor of 25 in pixel count. This also means that you can play with the DOF as you have taken 25 images with F10 with your F2 lens, assuming your lens is square. But its round so you lose some.
Yes, their 11 Mega rays don't really tell us much and is as much marketing blurb as anything, what is a ray (one photon wide?), how do they measure a ray's direction in practical terms, how many actual pixels etc...
 
I completely agree that this is rather untested in terms of actual market acceptance and I certainly wouldn't want to bet on the success or failure of it at this point. At the same time I can see exactly what they're trying to do with it, which I think a lot of people criticising various aspects of it are missing.
 
I do wonder why focus bracketing isn't implemented on some of the DSLRs
I assume its partly a user interface problem and partly it is because how auto focus works.

Its not so easy to set up the camera for focus bracketing. You have to choose the focus range. You have to match the aperture in some way to that range.

Moreover - I dont think auto focus cameras necessarily have any good notion on how they focus. They simply try to get the image sharp. Thats probably why there are nu useful EXIF info regarding focus distance. So - it might not be technically possible to bracket at all.

--
Roland

support http://www.openraw.org/
(Sleeping - so the need to support it is even higher)

X3F tools : http://www.proxel.se/x3f.html
 
Sure, but proprietary RAW formats don't make it a breach of the contract to, eg, use Lightroom on the files. They may make it inconvenient, but, eg, Nikon don't try to make it illegal to turn a NEF into a DNG.

I think what we are seeing is a camera business model we are not used to: Lytro is not selling you a camera, they are selling a website and software package of which the camera is a peripheral.
 
Sure, but proprietary RAW formats don't make it a breach of the contract to, eg, use Lightroom on the files. They may make it inconvenient, but, eg, Nikon don't try to make it illegal to turn a NEF into a DNG.

I think what we are seeing is a camera business model we are not used to: Lytro is not selling you a camera, they are selling a website and software package of which the camera is a peripheral.
Absolutely, due to a final low resolution (540x540 pixels).
 
Might be just the thing for surveillance video/photos!

The current ones are never clear enough to identify the suspect!

--
markeast
 
From the interviews and other information, this first version is quite preliminary.

With additional processing said to be coming in the next year, depth of field control and 3D output will become available.

One shot and you have any DOF,
No, you can't. You can have any DOF between that of the lens wide open and about 2 stops down for this production model, with about 4x decimation. The prototype, built from a medium format digital, could go almost 4 stops down.
Not sure why you're calling it decimation (a bit prejudicial), but assuming a Bayer sensor requires an even microlens imaging, 4x4 or 6x6 seems reasonable, so you can go from f/2 to f/8 or f/12 on the small sensor.
And the production camera, with a 35-280mm f11 equivalent lens, isn't going to deliver much in the way of shallow DOF.
Which seems contradicted by your below point of 0.24 feet DOF at max telephoto?
any focal point,
No. The focal point has to be in the DOF of the wide open lens. Now, that's pretty deep at the widest setting, but not at the longest setting. Look at the DOF at 10 feet...
  • 35mm f11 DOF = 55 feet, from 5.5 to 60.5 feet.
  • 280mm f11 DOF = 0.24 feet, from 9.88 to 10.1 feet.
Actually the focal point can be anywhere in the DOF of the (as you put it) "decimated" aperture, which is effectively f/44, so for
  • 35mm f44 DOF = infinite, from 2.3 to infinite feet
  • 280mm f44 DOF = 0.96 feet, from 9.54 to 10.5 feet (I can use DOFMaster too :) )
However, I wouldn't assume the preset focus distance is 10 feet. If I were designing it, I'd use hyperfocal distance and have focus change with zoom, so:
  • 35mm f45 DOF @ 3 feet = 116.7', from 1.5' to 118.2'
  • 280mm f45 DOF @ 180' = 3193.7', from 92.5' to 3286.2'
(DOFMaster didn't have f44, but I wonder how COC should be set for a plenoptic camera.)
You can refocus a bit to compensate for missed autofocus, but it's sure not "any focal point".
and/or 3D views.
A very unnatural "extruded" looking 3D, because a single point approach doesn't address the occlusion problem, as has already been discussed. If you've ever used a 3D paint program, you know the sort of ugly look it often gives.
Actually, the AsiaD video shows a parallax demo which demonstrates occlusion which is possible because they aren't taking from a single point, but across the microlens - effectively like having two sensors a microlens apart (admittedly a small base distance, but it is fairly effective in the parallax demo - I wonder if it isn't across the 17mm effective aperture).
--
Rahon Klavanian 1912-2008.

Armenian genocide survivor, amazing cook, scrabble master, and loving grandmother. You will be missed.

Ciao! Joseph

http://www.swissarmyfork.com
--
-- Please remove the Quote option!
-- Why can't you edit more than once???
-- How about switching to real forum software?
 
Actually it doesn't quite work that way.

It is more like the microlenses combine sensor pixels into NxN arrays and then they record images from multiple POV but at the same focal length and aperture. Then a lot of processing let's you compute effective DOF for aperture x N (so for the 35mm equivalent (guestimated) f11, you can compute f44)).

Due to Bayer demosaicing (and I assume off the shelf sensor plus don't reinvent the wheel) N probably has to be even, so not a 5x5 but either a 4x4 or 6x6 array.

They already mention their 11 "megaray" sensor can produce 1080x1080 (1MP) images.

Of course, the math works best with a 3x3 array, but then they'd need their own demosaic and their own color array.

--
-- Please remove the Quote option!
-- Why can't you edit more than once???
-- How about switching to real forum software?
 
Sure, but proprietary RAW formats don't make it a breach of the contract to, eg, use Lightroom on the files. They may make it inconvenient, but, eg, Nikon don't try to make it illegal to turn a NEF into a DNG.
Maybe
I think what we are seeing is a camera business model we are not used to: Lytro is not selling you a camera, they are selling a website and software package of which the camera is a peripheral.
We will see.

I recommend you though to read the fine print of software or gadgets you buy. Sometimes you can read real strange things there. Companies have a tendency to over protect themselves.

--
Roland

support http://www.openraw.org/
(Sleeping - so the need to support it is even higher)

X3F tools : http://www.proxel.se/x3f.html
 
This is more like using a Foveon sensor than just a different format - even if the RAW was provided, it is much different and much harder to produce a picture than just a demosaic.

The math is hairy and I bet they are on the Mac first because OpenCL was needed to process the images in a reasonable time.

--
-- Please remove the Quote option!
-- Why can't you edit more than once???
-- How about switching to real forum software?
 
Actually, they have an exact idea of where they are focusing - a phase detect sensor measures focus distance directly, and a contrast system knows it indirectly from the lens information.

Depending on the camera, however, they only have to be reasonably within DOF to call a picture focused.

--
-- Please remove the Quote option!
-- Why can't you edit more than once???
-- How about switching to real forum software?
 
You will get no argument from me that EULAs typically read as if they were written by beings from another galaxy. Why that is ... IMO, it has a lot to do with trying to use a nuts and bolts business model in a digital environment, and a lot to do with the psychotically deranged decision of the US courts some years ago to allow patents on the vaguest of ideas.
 
I do wonder why focus bracketing isn't implemented on some of the DSLRs
I assume its partly a user interface problem and partly it is because how auto focus works.

Its not so easy to set up the camera for focus bracketing. You have to choose the focus range. You have to match the aperture in some way to that range.

Moreover - I dont think auto focus cameras necessarily have any good notion on how they focus. They simply try to get the image sharp. Thats probably why there are nu useful EXIF info regarding focus distance. So - it might not be technically possible to bracket at all.
As I said it already exists in some bridge cameras.

It is dead simple to take a shot, adjust focus forward, take a shot, adjust focus backwards, take a shot. The only question is by how much to adjust, which can be left to the photographer to specify.

The down side for Canon would be if it reveals that many lenses are not focusing precisely.
 
It is dead simple to take a shot, adjust focus forward, take a shot, adjust focus backwards, take a shot. The only question is by how much to adjust, which can be left to the photographer to specify.
Hmmmm ... that was what I meant. "for the photographer to specify" - how?

NOTE that focussing series for Lytro simulation is NOT focus bracketing.
The down side for Canon would be if it reveals that many lenses are not focusing precisely.
--
Roland

support http://www.openraw.org/
(Sleeping - so the need to support it is even higher)

X3F tools : http://www.proxel.se/x3f.html
 

Keyboard shortcuts

Back
Top