Adobe develops Plenoptic lens

<p>Soon we could take a picture, and change the focus later!<br>

http://www.crunchgear.com/2010/09/22/adobe-plenoptic-lens-tech-lets-you-change-the-focus-in-your-images</p>

<p>Adobe has been talking about this for a year or two. Others have been taking about it even longer.</p>

<p>In 2005, a team lead byDr. Ren Ng at Stanford build up a plenoptic camera, made a bunch of shots, and got all sorts of media attention. In 2008, he went commercial and started up a company called "Refocus Imaging". After some press in 2008, nothing, dead silence, and their web site has essentially zero content.</p>

<p>The "dirty little secret" of all the lightfield, plenoptic, wavefront coding, etc. projects is that there's laws of information density that make the cameras essentially impractical with any camera sensor technology likely to be on the market in the next 5 years. It's literally a decade or more away. Here's a quote from the article you linked.</p>

<blockquote>

<p>The downside is vastly reduced image fidelity, since so much of the sensor is dedicated to more or less redundant data. He says it’s “on the order of” 25-50 individual cameras,</p>

</blockquote>

<p>"On the order of" is an understatement. The Stanford camera commits 192 sensor pixels to every output pixel. It used a 16mp 4000x4000 36x36mm MF back with a 9 micron pixel pitch, and a 125 micron pitch microlens array.</p>

<p>That netted them a 288x288 image. It looks OK in web sized animations. It probably even looks OK on a year 2005 projector, 1024x768 or lower, because all the attention is on the animation (oooh look! it's refocusing!) and not on the resolution.</p>

<p>I've (and worked) seen other math that says 100 could work well, or even 64. That could net a pretty tolerable 10mp image from "just" a 640mp sensor, and that's a pretty generous 1.16 micron pixel on full frame. You can't do it on anything smaller, because a 640mp APS takes the pixels down to light wavelengths, and you'd have to build a resonant aperture sensor (which is theoretically possible, but hasn't been done yet) and P&S sensors would be too far below a visible wavelength.</p>

<p>I doubt 50 or 25 is doable, regardless of what the Raytrix people are claiming (trust someone with "tricks" in their name).</p>

<p>Thanks, Joseph, for the excellent information! I agree that it is years away from commercial or consumer use. <br>

I have to say though that I find it exciting to learn that boundaries are being pushed, and that perhaps the industry isn't quite as "static" as I'd assumed.</p>

<p>On the way to a full range system, I would find one that just lets me touch up the focus very useful. This would require far fewer sensor pixels per output pixel I would imagine. I am thinking of shots where people are moving for instance. Under such circumstances I very often wind up with the focus on the nose tip or just behind the eyes. Just shifting that focus point slightly would dramatically increase my keeper rate.</p>

<p>Surely it would be more useful to develop Fourier Transform algorithms working on the RAW pixel data of <em>current, buyable cameras</em>? Who's going to shell out for a special, low res microlens camera, just so they can do one little trick with it? And if you're going to have a multiple lens camera, why not just split the sensor field and have 4 or more planes of focus captured as discrete and separate images?</p>

<p>In theory it's possible to restore focus on any out-of-focus image with a high degree of accuracy, provided the lens and sensor transform function, or point-spread function are well-defined. These transforms vary with the amount of defocus, so it should be possible to shift focus in much the same way as the nutty professor's microlens toy above.</p>

<p>If you'd been told that colour aberration fringing would be easily fixable, say 10 years ago, you'd have thought it was some far-fetched sci-fi fantasy, but now you can turn a cheap lens into an Apo version with ease. Better image processing is the way forward, not some hardware blind alley. And Adobe of all companies should know that!</p>

<p>Aaaaaahhhhhahahahahahahaha!</p>

<p>Sucker!?</p>

<p>The whole idea is ridiculous! It's like the BS about reading license plates from orbit. Can't happen. Doesn't happen. Sorry, but physics is physics, and you can't put together something that is impossible to put together, just like the BS they do on TV with "enhancing" video images to make someone's face appear from a blur. No real. FAKE! BS!<br /><br /></p>