First commercially available light field camera

[Dave Blackhurst]

OK, but they've already compressed a huge "concept" camera array into a small, relatively affordable, handheld, consumer ready device with acceptable output... at least for web and PC (er... Mac) display/use. I'd say they're doing rather well with whatever "physics" and "science" are involved...

One always has to remember that bees "shouldn't" fly and men aren't supposed to, but bees don't know any better, and man refuses to accept such facts.

For what they have, it's still pretty impressive, IMO, and a sign of interesting possibilities to come. Play with their samples and tell me they are not usable or effective in "scene capture"...

[David Heath]

Originally Posted by Dave Blackhurst

OK, but they've already compressed a huge "concept" camera array into a small, relatively affordable, handheld, consumer ready device with acceptable output... at least for web and PC (er... Mac) display/use. I'd say they're doing rather well with whatever "physics" and "science" are involved... ..

Oh yes, I take my hat off to them for what it is. And yes, great novelty to play with refocussing already shot images. But as you say - with "acceptable output... at least for web".

The applications are actually more than that, and it's easily forgotten that cameras in general are heavily used for industrial/scientific applications - not just "art" photography and filmmaking. Take a look, for example, of another plenoptic camera and it's application - Facial recognition - Raytrix GmbH . The idea is that for face recognition it's good to trade off two-dimensional resolution for depth. Same with many industrial uses. Don't think I'm dismissing the technology out of hand - far from it.

But where I disagree with previous comments in this thread is to then take a leap and see it as the future of general photography. Here, the resolution of these cameras falls way short, and even allowing for that, there would be huge problems capturing "11 megaray" images at video frame rates with current technology.

Yes, technology moves on, but you really have to have a decent grasp of how the system basically works to see just how big the problems are, and why they are unlikely to become mainstream any time soon - if ever. The Stanford team used a 16 megapixel camera, with a 90,000-microlens array in front of the sensor (each microlens covering about 175 pixels), to give a final resolution of 90 kilopixels. (About 320x240) To improve that substantially, you have to either:

Have far greater than 16 megapixel as base
Use a smaller no of pixels under the microlens - with consequent lessening of control

Certain other tricks are possible - but they have their own costs. And don't forget the data storage etc - in the original example you have to store data equivalent to a 16 megapixel image to get the 320x240 image. Imagine the file size corresponding to even such as a 6 megapixel final image!

[Dave Blackhurst]

A: Moore's law
B: My USB jump drive and all my memory cards have significantly more capacity than my first Hard Drive (30 whopping Megabytes!), and I have a couple TERAbytes of "local" storage.

Not sure I see the problem... although I'll admit with every new technology "toy", my computer always seems to need upgrading to deal with MORE 1's and 0's, faster... but all the technology just opens up cool new possibilities!

[David Heath]

Originally Posted by Dave Blackhurst

A: Moore's law

Not sure I see the problem...

That's what is meant by the differences between a technology limitation and a "laws of physics" limitation. Computer memory has been definately in the first category, but, as example, look at cars as an example of the latter. Their physical size is limited by the dimensions of the human body, and that leads to engineering constraints, let alone practicalties - they may (with increasing difficulty) be engineered to go faster, but the constraints come down to safety, ability to corner etc. Nobody expects them to get twice as fast year on year, nor give twice the mpg figure year on year. Not even close.

And one simple of law of physics which CANNOT be changed is the wavelength of light. That fixes an upper limit on pixel density, amongst other things. Just think of the maths. To go from about 320x240 to 1920x1080 is very roughly a 100,000 fold increase - and that's "only" to video HD resolution, not what is generally expected of a still camera. So, keeping other factors the same, that would mean a base chip resolution of about 1,600,000 megapixels!!!!

I'll let you do the maths (assuming the same chip size) as to what the inter-pixel spacing then becomes, and how it relates to the wavelength of light......... :-)

See the problem now........?