Photos you can refocus: Lytro promises camera within a year

This reminds me of the "we, real photographers, ..." LiveView discussions and decades earlier ones in other media about autofocus. :-)

This allows you to achieve
* Single-shot stacked focus (everything in focus)
* Single-shot, single-lens 3D (actual 3D depth map - not stereoscopy)
* Adjusting the focus after the fact.

Page 25 of the company's CEO's Ph.D thesis makes it clear what they are doing: http://www.lytro.com/renng-thesis.pdf

Necessarily, this method of capture trades off resolution for ability to "play" with the 3rd dimension and how finely. I.e. if you take a 20MP conventional sensor and use it for this application, if the fidelity of depth control is to be similar to what is indicated on the demo pic above, the image resolution would be somewhere around 1 megapixel.

This sounds like a non-starter; however if one were to take the compact camera or cell phone pixel density and put it on a full-frame sensor, one could counter the resolution problem.

Reading the comments below, it's obvious that most of you are completely misled.

Plenoptic imaging brings forth more than just "depth-of-field" control. It brings about the ability to change the perspective of the image after the fact, and opens the door to auto-stereoscopy. Changing the focus after the fact is a great feature, the purpose of which isn't to gain more DOF as most of you seem to be obsessed about: There is such a thing as creative use of limited DOF, and the ability to shift the focus in an image is also quite literally the ability to shift the point of attention — changing the character and subject of an image radically.

The best market for this may be traffic cameras shooting moving cars.

Things will get even more interesting when industry decides on a standard and couples light field photography to microlensed displays. Then there will be no need to process the images, the viewer's eyes and brain will do all the processing: refocusing, perspective correction, depth perception, etc.

The first steps have been taken for 3D TVs without the need for special glasses, the first steps have been taken for light field photography, now combine the two!

Wow!!! Very interesting!
So we won't have no more Front Focus or Back Focus problem to worry about...

It's like an automatic transmission. For someone who enjoys the art of photography, it's blasphemy. For someone who wants the best pictures of a moment, it's a no brainer.

I notice that in the sample given, the place where I click to refocus really doesn't correlate at all well with where the focus ends up getting put.

I'm excited about this. It may not work perfectly,and it may not be for everyone, but developments of this sort are certainly good for photography. There are so many talented photographers on dpreview.com and elsewhere, I can't wait to see what everyone can achieve with this camera.

Ultimately, most digital camera users are 'lazy' and want instant gratification.
This will likely mean that results will only ever be used to show what has just been taken, on the camera LCD, or posted on-line for others to play with.
Anything else and it turns every user into a RAW photographer, having to process all their pictures, and there is great resistance to that among the vast majority of users.
Indeed, to add choosing point of focus and perhaps depth of field to any adjustments of exposure, white balance and cropping makes the task of post-processing far too long-winded or intimidating for all but the keenest and I think the novelty might soon fade.
I'm certain it will have it's place, as with 3D cameras, but it won't be mainstream anytime soon.

Well , interesting development upon imaging side , but I am not seeing Nikon , Canon or anyone being threatened. It will still require a decent lens, a decent camera, and mostly this require a highly resolved sensor level capture to give us equivalent of todays.

The technology is useful, especially for many industrial / surveillance / scientific need. But its not going to replace good old photographic capture the old way.

Maybe it won't be this exact camera or company, but this concept, once the necessary hardware horsepower has been developed, could put Canon, Phase One, Nikon, Leica, and all the others out of business. It will finally be possible for photographic artists to achieve their vision without the hassles of switching lenses, and bad focus, and lens aberrations and all the other baloney we have to put up with. Friction-free photography is almost here.
Hurry Dr. Ng, hurry.

From page 60 of the thesis (http://www.lytro.com/renng-thesis.pdf) I understand that if each microlens covers m x m pixels and the image was taken at f/N then the image can be refocused anywhere in the depth of field of f/(N x m). For example, the image taken at f/4 and each microlens covers 8x8 pixels then the image can be refocused anywhere in the depth of field of f/32. This also means that spatial resolution is reduced by a factor of 256, say from 256 megapixels to 1 megapixel.

The tradeoff is that for each factor of 2 of loss in spatial resolution the "potential" depth of field increase corresponds with one step of aperture.

It's not that everything is in focus, it's that the selective focus can be edited. Of course, isn't that what you can do with Photoshop's lens blur filter and a reasonably sharp photo (say a 28 to 35 mm shot @ f4 or f5.6)? Point is, with a reasonably wide angle lens stopped down a bit, everything is in reasonable focus and it's not too hard to blur is.... This seems more like a gimmick than an innovation. Now, is this could be adapted to generate something where everything in a snap shot really is in sharp focus, that would be remarkable.

The technology is interesting . . .

But as a useful 'breakthrough" I would rank this a little above "Smile Detection." and well behind "Live View."
Camera makers are not even close to perfecting autofocus yet.

This is one of those "Holy Shit!" Moments. This completely changes photography. I honestly don't know how someone can say "it's just a matter of finding the right algorithm". This is major major stuff, a radical change from what photography used to be, form what we used to think pictures needed to be. Amazing.

Alas, the photo here is a cheat, but this really IS cool. If the data -- i.e. the photons -- are there, then it's just a matter of finding the right algorithm to make them coherent. Of course it's extremely complex in application, and Mr. Ng is a genius -- a new Edwin Land.

Blurb on their Website: "Lytro will debut the first light field camera for everyone. OK – you’re not everyone. You are a beautiful, unique snowflake. And you deserve an amazing camera...."

Love 'em already!

Although changing the point of focus is a nifty feature, this seems to have overshadowed something else - the camera lens can be fixed focus. This means the lens can be smaller and cheaper and also, importantly, there is no waiting for the camera to focus.

Notice how the Flash widget is actually presenting a number of discreet layers, and using a visual effect to "refocus" between these layers depending on where in the image you click.

For example, try focusing on the back wall. You'll not get the focal point any further away than the rear flask. Also, try refocusing by clicking on various points along the right hand edge of the workbench, where the holes are. You'll find that focus always moves to the rear flask, regardless of how far forward you click.

This is certainly an intriguing concept, and the science behind it is perfectly sound. But I'm not getting my hopes up until I can hold one of these cameras in my hands and see the results for myself.

you can read up on the technology at
http://graphics.stanford.edu/papers/lfcamera/lfcamera-150dpi.pdf

a down side is that the camera will have a fixed f number as the f number of the main lens and the micro lenses need to be the same (my guess is that the first model will be at f/4)
getting this to work with a zoom will require lots of tradeoffs

this is for the point and shoot market and not for the DSLR folks
seems that this could be adapted to cell phone type cameras as well