View Full Version : New Lytro camera
Dan Fleming
10-19-2011, 11:07 PM
I was pretty surprised not to see any discussion yet of the Lytro cameras (I only looked through the first couple pages, so this may be old hat). Check lytro.com out.
Now that you're back! Thoughts? Good, bad and ugly?
First off, a f/2 8x zoom (no specs on 35mm equivalence) is amazing for such a budget cam.
Second, image quality on their pics on site isnt great.
but the focus later ability is kinda neat.
Jordan
10-20-2011, 01:03 AM
Personally I think that the focusing afterward, though could certainly have some significant advantages in certain circumstances, as well as being "fun", isn
Dr Croubie
10-20-2011, 02:47 AM
I think i
They seem to be hitting the "instant" angle quite hard so far, in that there
I can see this easily adopted in my applications where dynamic and refocusing is critical like Dr. Croubie suggest security camera
neuroanatomist
10-20-2011, 04:58 PM
no idea what the number of pixels are,
...and they carefully don't say, possibly because it's not as simple as a count of pixels. But, my guess (and admittedly, there are a lot of underlying assumptions!) is that it's a 1/2.3" 14 MP sensor, meaning 1.4µm pixels. Because of the optics involved, I speculate that what you're actually getting out of the camera is the equivalent of a 3 MP image.
Here's my reasoning:
They have stated they use a 'standard sensor'. The sample images are all square, as is the LCD, but square sensors aren't 'standard' nor would they be cost-effective. So, they are almost certainly using a common 4:3 sensor (so they're throwing away some pixels, possibly for optical or computational reasons). The specs state, "11 Megarays" and while that's clearly not an SI unit of measurement, based on the technology 11 million rays means 11 million pixels. Assuming that's rounded up slightly (as is common), a 4:3 sensor would need to be a 14 MP sensor for 11 MP in a 1:1 aspect ratio.
Now, how big is the sensor? I estimated from the cutaway diagram, which appears to be to scale (seems plausible as the length and width of the diagram scale properly to the specified camera dimensions). I believe that the actual sensor is yellow, and the microlens array adhered to it is blue. One dimension of the sensor is ~4.6mm. Now, if I were selecting the diagram to use and intended to label the sensor, I'd want to show it as big as possible, I'd show the long dimension of the 4:3 sensor. A 4:3 sensor with a long dimension of ~4.6mm seems like a 1/3.2" sensor, and I can't find any that hit 14 MP. However, if we assume that the short side of the sensor is 4.6mm, that would indicate a 1/2.3" sensor. 14 MP 1/2.3" sensors (both CMOS and CCD) are used in well over 50 compact cameras, and are produced by many name-brand (Sony, Samsung, Toshiba) and off-brand companies, so they're pretty inexpensive.
Ok, so it's using 11 MP of a 14 MP sensor. Does that mean you're getting 11 MP of resolution? No. Because of the way a plenoptic camera works, you are essentially trading resolution for the ability to refocus. Reading over Ng's thesis, he's come up with a clever modification to mitigate that loss of resolution to ~50% linear, meaning 25% actual, thus 11 MP reduces to 2.75 MP.
Daniel Browning
10-21-2011, 12:19 AM
I can't really see any applications for this beyond a novelty/collector's item.
The biggest application for this is security cameras. One mounted on the side of the building, a few people unsavoury walk past, you can refocus and zoom on their faces individually to get better photos for IDing.
I don't think I understand. Why would the ability to refocus be superior to having everything in focus from the start (as most security cameras are now)? Are you saying that they walk so close to the security camera that they are out of focus and can't be identified?
wickerprints
10-21-2011, 04:03 AM
I'm surprised at how enthusiastic some people are about this camera, despite no information being given about the imaging resolution. Some idiots actually had the nerve to say the imaging resolution doesn't matter at all, because what's important is the refocusing capability. Would they say this if the imaging resolution was, say, 128 x 128 px? Yeah, because I would love to be able to refocus a tiny thumbnail-sized image. Resolution matters. It always has, and it always will. Anyone who thinks otherwise is delusional.
Even if we assume that the output image is 540 x 540 px, it still presents a huge tradeoff--do you take the picture with a Lytro camera just so you can play with focus, or do you take a picture with a normal camera so that you can make prints, crop, adjust color, display full screen, etc.? A point-and-shoot 14 MP camera could give you much finer details. Modern AF systems are so fast and so good these days, and the sensors are so small, that missed focus is just not a problem for most consumers. It's been repeatedly claimed that because this camera isn't intended for serious photographers, that it's somehow okay if it has low resolution. Well, I hate to break the bad news, but even casual consumers know that resolution is important, and they value being able to get more details, better color, and better performance out of their cameras. It's not just the serious photographers who care about these things.
Yes, the idea is a novelty, and yes, I would like to be open about the creative potential of the technology. But the fact that Lytro intentionally refuses to disclose something as simple, concrete, and important as the size of the image, means that it must actually be very low. When I pointed this out, someone retorted that there was no reason for Lytro to weaken their product marketability by mentioning it. Well, that presumes that there are a number of consumers who would want to buy this product who wouldn't if they knew that it will only produce web-sized images. Then such an omission is a deception being perpetrated upon them. They have a right to know before they buy. If you argue that consumers would be unjustly disenchanted by a low spec, but in fact would love the camera if they could just get over the numbers, then they STILL have the right to know. If the camera is as good and worthy as Lytro believes it to be, and if this feature is worth the resolution hit, then the product will succeed on its own merits. The bottom line is that it is never a benefit to the consumer to deliberately withhold information about what the product can do. If it truly deserves to do well, it will do well. And if it deserves to do poorly, don't trick people into buying your snake oil. Can you imagine how such tactics would play out if, say, Canon did it? Imagine what kind of reaction people would have if Canon said, "our new EOS 1D X has a new '50 megaray' sensor!" and refused to explain what that meant or how it translates into pixel count.
Lytro initially gave the excuse that the resolution of their camera was still undecided, which is why they didn't give any details. Well, now that excuse is gone, and now they've coined some new terminology that is understood only by them. Furthermore, anyone who has read the research knows that plenoptic imaging's #1 obstacle is spatial resolution. So to be coy about this information is incredibly suspect. If Lytro--and by association, Ren Ng--had been up front and honest from the start, I wouldn't be condemning this business venture. But at every possible turn, he has chosen to do the dishonorable thing, and that's why I hope this product fails miserably.