Re: One day there might be a Lytro video camera
 

But beware of what they DON'T say. The basic principle behind Lytro is the ability to trade resolution for refocussability etc, and yes, it may do what it says - to a point.

It does it - but as well as the much talked about resolution drop (final image compared to basic sensor) it's not infinitely refocussable as a basic lens would be. Think of it moving in steps, rather than a continuous smooth adjustment. You may be able to tap areas on sample images and get it to refocus between them - but you may not be able to focus between two planes, for example.

How bad is that? Well, the more you trade off resolution the smoother the post-focussing will be - one improves at the expense of the other. The only way to improve one without the trade off is to increase the number of individual sensors on the chip. In this respect it's a bit like taking a photo of the audience from the stage. You may want to set the point of focus on row 12 (which you could with a normal camera) - but with Lytro you may find you can post set it to row 9.... or row 15! But not precisely on row 12. And because the basic image resolution is relatively low (about 1 megapixel?) then in a way it helps, because you are less able to see such effects. But try to use it as a post tool on feature film quality material and you'll need a HUGE number of basic sensor elements. Even if possible, (and see next paragraph) the data storage requirements would be mindblowing - if you think 4k is bad......

Their CEO talks about Moore's Law in the quote - but unfortunately, that doesn't really hold up here. There is nothing he can do about the wavelength of visible light, and that limits the size of photosites on any chip, and hence the maximum possible number on any given size. The problem is one of the laws of physics - not technology or engineering. And we all know what Scottie had to say about that....... :-)

The article also talks about 3D, and whilst it's true light field technology gives some ability to "look around" an object, then again it's limited. I understand that the best you'll be able to do is imagine a cut out hole the diameter of the lens, and the effect will be as if you were looking through the hole and moved your eye around top-bottom, left-right. The perspective will change - but by a limited amount. To expect light field technology to revolutionise film production, being the technology that replaces current 3D hardware and on set focus pulling simply will not happen.

And light-field video cameras DO exist already - even if not via Lytro. See http://www.raytrix.de/tl_files/downl...rix_Slides.pdf for examples of what they are currently being actively used for. But their real use is coming about in industrial applications - not photography in it's more artistic sense. Think say of face recognition. You may not need huge resolution - but for such an application, one megapixel images with a depth map of the face may be vastly better than conventional 2D images at higher basic resolution. Think of medical uses.

Re: One day there might be a Lytro video camera
 

Awesome post, especially with the info on the stepping limits for refocusing an image.

I had noticed that you never saw them specifying what size the final Lytro images actually were. It had to be bad if they were hiding it that much. For a moment I was thinking they might be able to make use of a 4K sensor chip. :-)

Andrew

Re: One day there might be a Lytro video camera
 

Just to clarify, what I said above somewhat oversimplifies the situation to try to get the issues across, but the basic truth is that for a basic sensor with a given number of individual photosites, what light field technology does is sacrifice resolution for depth information. The more accurate you need the depth information, the more basic resolution you have to sacrifice. You have to rob Peter to pay Paul.

To keep even HD (say 2 megapixels) final resolution, with sufficient depth information for proper post focussing, you're likely to need at least 100 million photosites, which means a likely necessity for chips bigger than s35, and just think of the data rate if it's to read that at even 24 times a second!

But the real point is that there already are light-field (albeit not from Lytro!) video cameras. It's just that their resolution and performance are nowhere near good enough for "film-making". That's fine - they are aimed at industrial use and do that very well. But scaling the technology up to a level for film-making would be a huge challenge, if not impossibility - and as said before, it's more for "Laws of Physics" reasons than technology. That's the real problem.

Re: Light Field Technology!
 

Lytro's CEO throws in the towel, sadly:

"The goal now is to reshape VR, not consumer cameras."

The Lytro camera ceases to be...

Lytro CEO admits competing with Canon, Nikon, and smartphones was a losing game | The Verge

and

https://backchannel.com/war-stories-...2a6#.ugm6upj8x

Re: Light Field Technology!
 

And then steps back in the ring with a truly professional camera:


So it seems like they've just dropped the consumer market, which was always a questionable one for the technology, and decided to focus on the professional market where they can potentially achieve what the technology has always promised.

More info on the camera here:

http://techcrunch.com/2016/04/11/lyt...ative-freedom/

Re: Light Field Technology!
 

And they're even talking about resolution received from the camera! My, how things have suddenly changed.

Andrew

Re: Light Field Technology!
 

Not really. All they really refer to is a basic "755 megapixel sensor", and from that someone has inferred "40K" as resolution. Do the maths and if you assumed 16:9 aspect ratio, it gives a horizontal photosite count of about 36.6K horizontally - which is where I assume the "40K" comes from.

But the whole point of the Lytro technology is that it trades resolution for other features (such as refocusing). So the final image resolution will be much less than 36K, the amount depending on the fineness of the "refocussability" etc. But it should certainly be up to at least 2K for video, probably even 4K, yet still have enough in hand to provide reasonably seamless post focussing. So far so good.

But ..... look at the figures again. I don't see any mention of sensor size? We're talking about the horizontal count going up roughly by a factor of about 9 compared to a standard s35 4K sensor, and it being read at 300fps. Either the photosite size must therefore decrease a lot, or the sensor size must go up. To keep the same photosite size, you're looking at a sensor of around 21x12 cms (wow! :-) ), and even if each was shrunk to only a quarter of the s35 4k size (so 1/16 the area), it's still about 5cmsx3cms in total. And if you're reading at 300fps, small photosites aren't a good idea.

Yes, it's possible, but even before thinking about the huge data requirements, expect it to be very, very expensive, and don't expect a zoom lens.

Certainly, don't expect it to be the future of general cinematography. (Not unless refocusability is worth more to you than zoom lens etc, before even thinking about the cost.)

The real question then is less will it work (I'm sure it will) - but will the additional functionality be worth the cost? (The article describes "This camera is essentially military-grade technology") Will it be simply more cost efficient, even for VFX work, to just use current techniques? Not as elegant maybe, but overall far more cost effective?

It's worth noting that they claim:
Well, not really. Such a camera is only capturing the rays that impact the camera lens, albeit the area at full aperture. And that becomes significant when you consider the 3D aspects Lytro refer to. It's quite true that such a light field camera can produce a more "true" 3D effect than a stereoscopic camera - but the limits of such are similar to looking at a real world scene through such as a hole in a fence. Perspectives will change somewhat depending which part of the hole you look through - but the limits will be set by the edges of the hole. In this case, the diameter of the lens corresponds to the hole, and limits the amount of "looking around" objects that's possible.