Can't find any information anywhere on the resolution of a Lytro photo. Not even on their website. That raises a red flag. What in the world is a 'megaray' and more importantly how do you compare that to megabytes for the purpose of estimating picture quality?
I'm not sure how they calculate megapixels from rays. My understanding is that post processing of the light field data (rays) yield the final image, but that better algorithms can improve the image. Rays aren't pixels; they are used to "synthesize" the pixels that are rendered on the target image plane. It's sort of like a holographic film: cutting the film in half doesn't delete half the recreated hologram; it only makes the existing hologram less detailed.
to answer my own question about size of the physical sensor, looks like it a 1" sensor (2.7x 35mm crop factor). Which would make the 30-250mm lens (35mm equiv. ) physically 11mm-93mm. It run Android on a Snapdragon processor. 2D resolution seems to be 5MP, not 1MP or 4MP as previously stated.
This is very much a niche product. With a fixed f/2 aperture, most photos are going to have significant swaths that are out of focus when the main subject is in focus. To see everything in focus requires interaction and can't be accomplished all at once in a single snapshot. Judging from characteristics of the original model, at 40 megarays the new model will have an effective resolution under 1 megapixel (or under 2 megapixels interpolated).
Isn't the point of light field that nothing is out of focus?
With a fixed f/2 aperture, most photos are going to have significant swaths that are out of focus when the main subject is in focus.
I think that is the whole point of how they do selective focus. I'm going to assume that "40 mega-array" means there is an array of 40 (1 MP) sensors stacked over the top of each other. Each capturing at slight offset focal lengths. You can have a depth of field based on a single focal length with a shallow depth of field, or combine some/all of the other 39 focal length captures to give yourself a large effective depth of field.
That seems very unlikely since it would be simple sensor focus stacking, not light field imaging. The rays in question should be the directionally-resolved rays separated at each image pixel location by a microlens and recorded on the actual sensor. That implies that the actual recording sensor (under the microlens array) is 40 MP, which is quite reasonable. The final iamge has the same pixel dimensions as the microlens array, and so if the new camera is using the same number of discrete rays per image pixel, then this camera should produce around 915 x 915 images.
Imagine if Apple could get an exclusive deal to put this tech in iOS devices. It could blow away all those 40 megapixel cameras. Even just on the gimmick of refocusing after taking the shot
The more I read, the less I worry about resolution.
And as I stated earlier, i.e., "If the product meets the initial interest, I won't hesitate to get one."
I'm not impressed by the fluff piece the Verge wrote. I just hope that the photo they had as an example wasn't the full Rez one,though I suspect it is. It was very fuzzy. Low Rez is low Rez. There is nothing you can do about that. It would have been more telling if the subject wasn't a piece of what exactly? I would have rather seen a photo of something recognizable.
Why not? “Focus after the fact” is a great innovation. I don’t mean lytro the company, I mean lytro’s technology.
Because right now, and I've seen no evidence so far that this will change, you have to divide the resolution of the sensor by 16 times to get the resolution of the picture. That means that this 40MP sensor gives a resolution of just 2.5MP. I don't know about for you, but that's way too low for anything other than Facebook pictures. In order to get a decent resolution image, you will need a sensor of at least 150MP, for a result of 9.373MP. How far in the future will we be before we can get sensors with that resolution? And even if we can, lenses won't be able to deliver that resolution in any decent sized sensor, because the light circle will be smaller than the wavelength of visible light.
There are practical limitations here that have to be looked at. So for low Rez, it's fine, but for a high quality picture, resolution plays an important role. And we're not talking about a megapixel race here, because in practice, everyone has reached the 20MP+ level in quality cameras. But they're doing that with 20MP sensors. To equal that, we would need a sensor of over 300 MP, which for anything other than extremely expensive photographic purposes will be well out of the price range of anyone. And that would have to be a very large sensor with a correspondingly large, expensive lens.
So I don't think this is the future of photography any more than stereo photography was thought to be going back to the late 1800's, and every few decades after that. And I'm not old fashioned either. In 2000, I wrote an article for Kodak's monthly magazine Lab Notes, which was sent to all the photo labs that used their products, that by 2010, film would effectively be dead. Kodak told me that they had a greater response to that article than any other, mostly negative. But, two years later, they said the same thing. I was also one of the first to have a commercial photo lab go digital, when in 1988, I bought the Crossfield System, for what was then a whopping $250,000. That consisted of a drum scammer, a Mac IICi, and pre Photoshop software. My partner thought I was crazy.
4 MP no matter you you compose the picture isn't really that bad. Beyond that the ability to reset focus and depth of field can be a wonderful thing in a fast paced setting or an environment with a lot of noise. I've had more than a few pictures ruined because the autofocus system decided to lock to a twig at the time I pushed the release. Or whatever it does that I didn't want it to do, the fact is there are many times when you want to recompose after the fact because you had little time to get it right at picture taking time.
I'm not sure how they calculate megapixels from rays. My understanding is that post processing of the light field data (rays) yield the final image, but that better algorithms can improve the image. Rays aren't pixels; they are used to "synthesize" the pixels that are rendered on the target image plane. It's sort of like a holographic film: cutting the film in half doesn't delete half the recreated hologram; it only makes the existing hologram less detailed.
Decide "mega rays" by 16 to get the actual resolution. In this case, it's 2.5MP.
to answer my own question about size of the physical sensor, looks like it a 1" sensor (2.7x 35mm crop factor). Which would make the 30-250mm lens (35mm equiv. ) physically 11mm-93mm. It run Android on a Snapdragon processor. 2D resolution seems to be 5MP, not 1MP or 4MP as previously stated.
Imagine if Apple could get an exclusive deal to put this tech in iOS devices. It could blow away all those 40 megapixel cameras. Even just on the gimmick of refocusing after taking the shot
Not even close. I guarantee if you have a good quality image from a 5S, and the same image from this, you would prefer the 5S one.
Focus after the fact is a gimmick in the same way that autofocus before the fact is a gimmick.
I know you were joking. Focus after the fact would be a great thing if all the other factors were equal, but they're not. If you want a 4x5 image blown up to 8x10, then you'll be happy.
But after the fact focus doesn't make up for a soft, low detail image. And, most importantly, most images are out of focus by just a little bit. If the image was hi Rez, and it could be refocused, that would be great! But if the focussed image is softer than the out of focus portion you care about in the higher Rez photo, I don't see the big deal there. And truly, what percentage of your good photos are SO far out of focus that you would be willing to give up all of that image quality for the ability to sharpen those rare ones, most of which likely aren't very good anyway?
Comments
I'm not sure how they calculate megapixels from rays. My understanding is that post processing of the light field data (rays) yield the final image, but that better algorithms can improve the image. Rays aren't pixels; they are used to "synthesize" the pixels that are rendered on the target image plane. It's sort of like a holographic film: cutting the film in half doesn't delete half the recreated hologram; it only makes the existing hologram less detailed.
to answer my own question about size of the physical sensor, looks like it a 1" sensor (2.7x 35mm crop factor). Which would make the 30-250mm lens (35mm equiv. ) physically 11mm-93mm. It run Android on a Snapdragon processor. 2D resolution seems to be 5MP, not 1MP or 4MP as previously stated.
http://www.dpreview.com/news/2014/04/22/lytro-announces-illum-light-field-camera?utm_campaign=internal-link&utm_source=news-list&utm_medium=text&ref=title_0
This is very much a niche product. With a fixed f/2 aperture, most photos are going to have significant swaths that are out of focus when the main subject is in focus. To see everything in focus requires interaction and can't be accomplished all at once in a single snapshot. Judging from characteristics of the original model, at 40 megarays the new model will have an effective resolution under 1 megapixel (or under 2 megapixels interpolated).
Isn't the point of light field that nothing is out of focus?
That seems very unlikely since it would be simple sensor focus stacking, not light field imaging. The rays in question should be the directionally-resolved rays separated at each image pixel location by a microlens and recorded on the actual sensor. That implies that the actual recording sensor (under the microlens array) is 40 MP, which is quite reasonable. The final iamge has the same pixel dimensions as the microlens array, and so if the new camera is using the same number of discrete rays per image pixel, then this camera should produce around 915 x 915 images.
Focus after the fact is a gimmick in the same way that autofocus before the fact is a gimmick.
I'm not impressed by the fluff piece the Verge wrote. I just hope that the photo they had as an example wasn't the full Rez one,though I suspect it is. It was very fuzzy. Low Rez is low Rez. There is nothing you can do about that. It would have been more telling if the subject wasn't a piece of what exactly? I would have rather seen a photo of something recognizable.
Because right now, and I've seen no evidence so far that this will change, you have to divide the resolution of the sensor by 16 times to get the resolution of the picture. That means that this 40MP sensor gives a resolution of just 2.5MP. I don't know about for you, but that's way too low for anything other than Facebook pictures. In order to get a decent resolution image, you will need a sensor of at least 150MP, for a result of 9.373MP. How far in the future will we be before we can get sensors with that resolution? And even if we can, lenses won't be able to deliver that resolution in any decent sized sensor, because the light circle will be smaller than the wavelength of visible light.
There are practical limitations here that have to be looked at. So for low Rez, it's fine, but for a high quality picture, resolution plays an important role. And we're not talking about a megapixel race here, because in practice, everyone has reached the 20MP+ level in quality cameras. But they're doing that with 20MP sensors. To equal that, we would need a sensor of over 300 MP, which for anything other than extremely expensive photographic purposes will be well out of the price range of anyone. And that would have to be a very large sensor with a correspondingly large, expensive lens.
So I don't think this is the future of photography any more than stereo photography was thought to be going back to the late 1800's, and every few decades after that. And I'm not old fashioned either. In 2000, I wrote an article for Kodak's monthly magazine Lab Notes, which was sent to all the photo labs that used their products, that by 2010, film would effectively be dead. Kodak told me that they had a greater response to that article than any other, mostly negative. But, two years later, they said the same thing. I was also one of the first to have a commercial photo lab go digital, when in 1988, I bought the Crossfield System, for what was then a whopping $250,000. That consisted of a drum scammer, a Mac IICi, and pre Photoshop software. My partner thought I was crazy.
It's 2.5MP.
Decide "mega rays" by 16 to get the actual resolution. In this case, it's 2.5MP.
2.5MP.
Not even close. I guarantee if you have a good quality image from a 5S, and the same image from this, you would prefer the 5S one.
I know you were joking. Focus after the fact would be a great thing if all the other factors were equal, but they're not. If you want a 4x5 image blown up to 8x10, then you'll be happy.
But after the fact focus doesn't make up for a soft, low detail image. And, most importantly, most images are out of focus by just a little bit. If the image was hi Rez, and it could be refocused, that would be great! But if the focussed image is softer than the out of focus portion you care about in the higher Rez photo, I don't see the big deal there. And truly, what percentage of your good photos are SO far out of focus that you would be willing to give up all of that image quality for the ability to sharpen those rare ones, most of which likely aren't very good anyway?