Mobile Photography Set For Major Quality Bump With Sony's 48-Megapixel Sensor (newatlas.com) 112
Smartphone camera sensors and lenses have to operate in a very tight space, but they continue to close the gap on full-size digital cameras year after year. Sony's new IMX586 sensor boasts a 48-megapixel resolution, the highest yet for a mobile sensor, and should be coming to a phone near you soon. From a report: That increased resolution shrinks the pixel size down to 0.8 microns, which would usually lead to lower sensitivity and poor light collection. However, thanks to some smart technology called a Quad Bayer array -- where neighboring pixels are intelligently combined -- Sony says the effective pixel size is 1.6 microns. The bigger the pixel size, the better the light capture and low-light performance. In comparison, the Google Pixel 2 -- one of the best photo-taking phones on the market right now -- has a camera with a 1.4-micron pixel size. On paper, that means Sony has managed to produce a sensor that combines a huge amount of detail with excellent light capture and low noise levels as well. We'll have to wait until the sensor is actually on the market to know for sure, but the signs are good.
wrong metric (Score:5, Insightful)
If you blow up your photos to the pixel level, you'll find that it's not the pixel count that's making them look bad, it's the pixel-to-pixel noise and compression and color fringing, for example.
We don't need 48 MP taking up space on our phones and hard drives. For camera phone lenses, compressions, 24 MP is already enough. Anything more than that (if photography is your livelihood for example) and you should be relying on a DSLR.
Re: (Score:2)
We need to start a storage arms race. I could easily use PB HDDs but we're barely advancing on that front. Phones should start at terabyte storage, not laptops.
Re: (Score:3)
The laws of physics haven't changed appreciably that I know of, have they...?
Re: (Score:3)
Re:wrong metric (Score:4, Interesting)
Re: (Score:2)
You missed addressing the GP's point directly too: It's also good for noise since noise is distributed randomly across pixels. Sure the signal to noise ratio on individual pixels is better for a larger pixel, but for smaller ones the noise reduction algorithms work in a far more visually pleasing way.
Re: (Score:2)
Increasing the resolution of a low MP sensor is extremely difficult. Please note that I did not say it is impossible. it *is* possible to increase the resolution but it is a complex, time consuming, and sensitive process. Starting with more pixels is a much better option!
You have obviously not seen a detective show since the 70's. Every one of them has a sequence where they blow up a grainy, night time security camera video in order to read the washing instructions on the inside of someones underwear.
Re: (Score:3)
We don't need 48 MP taking up space on our phones and hard drives.
I don't think you've got the right metric there either. The right metric should be: "If I jpeg-encode a 48MP image to X number of bytes, vs jpeg-encode a 24MP image to the same number of bytes, which one produces the best image?" (for various values of X, I guess from about 0.5mb to 8mb)
I'd assume the 48MP one will produce the best image -- because we haven't pre-emptively thrown away information at the sensor stage, and so we've left it to the encoder to decide which information to throw away to achieve th
Re: (Score:3)
I'd assume the 48MP one will produce the best image -- because we haven't pre-emptively thrown away information at the sensor stage, and so we've left it to the encoder to decide which information to throw away to achieve that filesize, and the software can make a better decision about what to throw away.
That's a very questionable assumption, because if you are encoding a 48MP JPEG rather than a 24MP JPEG, you are still forcing the format to store information for twice as many pixels, whether or not there is meaningful detail in those pixels. Yes, it's compressed, but just go ahead and do this experiment for yourself - an aggressively compressed, high-resolution image will probably display noticeable compression artifacts, while a barely compressed image at half the resolution will probably look just as goo
Re: (Score:2)
That's a very questionable assumption, because if you are encoding a 48MP JPEG rather than a 24MP JPEG, you are still forcing the format to store information for twice as many pixels, whether or not there is meaningful detail in those pixels.
I don't know about jpeg compression in particular. But if compression is done with gabor wavelets then it doesn't need to store information for twice as many pixels. Nor fractal compression. And if you encode audio in the frequency domain then you don't need to store information for twice as many samples.
Yes, it's compressed, but just go ahead and do this experiment for yourself - an aggressively compressed, high-resolution image will probably display noticeable compression artifacts, while a barely compressed image at half the resolution will probably look just as good as the original, for realistic viewing conditions.
I did do the experiment myself, which is why I'm posting! I digitized all my old photo negatives. My target filesize was 3MB/photo. I spent a day just experimenting with compression, and determined that 9MB
Re: (Score:2)
The cellphone lens has been the bottleneck of its photographic ability for a long time now, even 10MP is a generous estimate. Enlarge any part of a cell phone photo and you'll see a mosaic effect caused by the lens limiting detail long before the pixels do.
Do they? (Score:1)
We all know that that's what the vast majority of consumers can understand, but for anyone caring about the details, it's relatively meaningless as a comparator.
Do people actually care?
Sometimes, I think marketers care about things because they think other people care and those people think other people care. I don't. I never look at the camera specs. Frankly, anything over 5MP is more than adequate for a phone.
Real estate. Every real estate agent spews the same thing, "The house and interior paint needs to be neutral colors yada yada yada yada..."
I asked a real estate agent once, "Whenever someone buys a house, the first thing they do is redecorate. Is there a
Re: (Score:2)
Actually a medium format (120mm?) sensor is even better. I don't know what the next size up would be but that would probably be even better yet (assuming pixel density stays relatively the same.
Just as in film the larger the sensor (film) size the better the end results.
Re: (Score:2)
Re: (Score:2)
Since the 70D came out, it's not clear full frame is even needed for low light. I assume Nikon has made similar advances.
Re: (Score:1)
Re: (Score:3)
We don't need 48 MP taking up space on our phones and hard drives.
And that's exactly why this is happening.
What Joe Consumer knows is that his 64 hippobyte iPhone filled up real quick, and he needs to buy a 128 rhinobyte one as soon as his contract is up. Or sooner.
This is all about up-selling devices and cloud storage.
Re: (Score:2)
As the summary points out, the higher MP count helps with low light performance and HDR. The smaller the individual pixels on the sensor the less light falls on them, so ideally you want the largest possible pixels. That's why DSLRs use much, much larger sensors.
One of the reasons why the Pixel 2 has the best low light performance of any phone is that it has an unusually large sensor. What Sony have done is create a sensor with small pixels that can be combined 2x2 to act as a 1/4th resolution single pixel.
Re: wrong metric (Score:2)
Re: (Score:2)
THANK YOU! (Score:1)
Re: (Score:2)
If you blow up your photos to the pixel level, you'll find that it's not the pixel count that's making them look bad, it's the pixel-to-pixel noise and compression and color fringing, for example.
On my 15MP camera, I am limited by coma and chromatic aberration before sensor resolution.
Megapixels aren't quality (Score:2)
Re: (Score:2)
Too much is alway better than not enough!
Unless 'not enough' triggers 'div by 0' (e.g. not enough children, not enough Clintons in white house etc, YMMV).
Re: (Score:1)
No shit? No one was saying such a thing. Hence why the summary goes into how Sony also added improvements to light collection to talk about how the sensor is better.
Re: (Score:2)
Re: (Score:2)
So I'm really surprised they can pack 48 MB efficient pixels on those small camera sensors.
Re: (Score:2)
https://www.dxomark.com/huawei... [dxomark.com]
Not necessarily, but it can be.
The Huawei P20 Pro scores the highest of any phone on dxomark... 40MP sensor, but averaged 2x2 with averaging to take great night shots for an apparent 10MP output. Apparently the 2x2 binning of smaller sites produces superior results to larger sites without the averaging.
I would guess this 48MP phone sensor would be the same deal... 2x2 averaged for 12MP output.
Sam
Re: (Score:2)
can the lenses keep up? (Score:5, Insightful)
The ability of the lenses to provide a sharp image at that (actual) pixel density is going to need some serious optical design. The sort that usually costs more than the whole phone.
Why not just use the quad bayer array with fewer pixels and give mobile phones an actual low light capability for once.
Re: (Score:3)
Re: (Score:2)
To be fair, the ability to focus within an inch and that narrow depth of field does offer some lovely photographic options.
Macro lenses can match the depth of field but not at that wide an angle.
But the pixel count and density aren't relevant to the depth of field. You could get lovely RAW images from a 16-20 megapixel sensor and give the lens far less work to do.
Still, create a lens that does f0.5 with any level of optical resolution and you'll be a rich hamste.
Re: (Score:2)
For interesting effects I have been known to stick a 17mm fisheye on up to 19mm of extension tube. A reverse mounted
Re: (Score:2)
Yeah, guessed you were messing. I just want a f/0.5 lens :) Also, diffraction is the term I've been trying to remember all evening, ta.
I'm too lazy to do much macro even with a macro lens. When the DoF is that fine you end up needing to focus stack and that's generally a pain in the arse.
Re: (Score:2)
Re: (Score:3)
https://petapixel.com/2013/08/... [petapixel.com]
TLDR: Sold for $80k, in the 60s, and doesn't work. Zeiss being silly, mocking the trend of the time.
Re: (Score:2)
unfortunately i will not be one of the 2 people that could maybe afford that lens...
Re: (Score:2)
You do realize that with the Quad Bayer array the camera's resolution drops to 12MP with ~1.6um sensor cells, right?
If released now it would be competing against an iPhone 8 at 12MP with ~1.2um sensor cells.
So just how much additional resolution do you think that a consumer would be willing to sacrifice in order to obtain improved low light capability? Just how large do those sensor cells
Re: (Score:2)
You're assuming that each pixel is represented exclusively by four other pixels.
I can't be arsed reading up on it, but that does feel unnecessary. Each pixel can capture its own light, be informed by the four pixels surrounding it and also contribute information to those.
As for commercial viability, consumers like big numbers. 48 mega pixels sounds so much better than 12.
Re: (Score:2)
No, I've read up on it [gsmarena.com], and you're projecting your tendency to assume onto me.
Re: (Score:2)
Perhaps you could have helped avoid confusion by mentioning that image capture can occur in two modes: Good light, at 48MP and less good light, at 12MP.
I'd still need to read more about the quad bayer array to properly understand how they're binning photons and computing pixel colour and intensity from that. To be fair sensors are something Sony do well.
Re: (Score:2)
The summary and TFA did that:
And yes, I did that:
"[N]eighboring pixels are intelligently combined" and "drops to 12MP with [2x sensor dimen
Re: (Score:2)
What lie?
48MP or 12MP wit/h high sensitivity. Yes.
Bingo. Specs report the detector resolution, not an "effective color resolution" after discounting for Bayer patterning. So where's the lie? Did I say "effective" or "color resolution" anywhere but here?
Re: (Score:2)
Why not just use the quad bayer array with fewer pixels and give mobile phones an actual low light capability for once.
Or better still, why not up to 48Mpxl and then us signal processing to eliminate the randomly distributed noise across the pixels to give mobile phones an actual low light capability ... while also providing the ability to give other useful features like better HDR.
Re: (Score:2)
Primarily because random noise is difficult to differentiate from actual image content.
Photograph the sky at night. Is that noise or a galaxy?
Wait a bit longer, count more photons hitting the sensor.. it's a galaxy. What if you don't want to wait? How do you get those extra photons?
Give them more sensor to hit. Larger pixels.
This is why larger sensors have lower noise in every sensor generation, and why you can't buy a full frame camera with the same pixel density as a mobile phone - scale the Samsung S9 se
Re: (Score:3)
Funny you should mention Galaxies. But not so funny how you're comparing the two.
Photon counting vs producing a photograph and noise reducing the results are two very different statistical processes. The benefit of the galaxies is they don't move allowing us to reset the nose floor between identical exposures thus statistically eliminating the noise of the sensor as you go. Just doing a longer exposure doesn't help you much determine what is Galaxy and what is a photon hitting the sensor.
Then there's the ac
Re: (Score:2)
The benefit of the galaxies is they don't move
That's wrong on pretty much every level.
allowing us to reset the nose floor between identical exposures
I like my photographs to work from a single exposure. A bird wont take a fish from water multiple times in a row to allow you to capture multiple identical exposures and eliminate the noise as the difference between them.
Just doing a longer exposure doesn't help you much determine what is Galaxy and what is a photon hitting the sensor.
The photons come from the fucking galaxy. A longer exposure does indeed allow you to receive more light from the galaxy, or star, or reflection of light off a sea eagle in its dive.
Of course, that sea eagle is moving quite quickly, which is why bigger
Re: (Score:2)
That's wrong on pretty much every level.
Wow. Just wow!
I like my photographs to work from a single exposure.
So do I. You're the one who started talking about galaxies. Something that precisely isn't imaged with a single exposure.
Oh look. The full frame sensor has discernably less noise than the APS-C sensor, despite both being 24MP.
Yep it did. You're not listing quantum efficiency, you're listing noise and seemingly ignoring half my post while you rave about hardware.
My camera can use sub-pixel shifts in the sensor too but that's still fuck all use with moving subjects
Ahhh exactly! Now we're talking the same language. So in order to get the same benefits as we discussed when you started talking about galaxies we can do something very simple: Record at a resolution higher than the diffraction limit of th
Re: (Score:2)
What you're failing miserably to comprehend is that all the signal processing, image processing, manipulation of the data captured from the sensor is entirely fucking irrelevant.
You can do all of it the same way on data captured from any sensor. It's a constant.
The variable becomes the pixel size of the sensor, and my point is that the larger this is, the more light it can capture, and the better quality image you will get as a result.
Try to understand what I'm talking about
I understand quite well enough. You're a megapixel count queen and think
Re: (Score:2)
What you're failing miserably to comprehend is that all the signal processing, image processing, manipulation of the data captured from the sensor is entirely fucking irrelevant.
You can do all of it the same way on data captured from any sensor. It's a constant.
You can do it anyway. However when you do it on resolutions that at lower resolutions you start to actively clobber parts of your data since these algorithms rely primarily on looking for differences in very small spaces. The only time what you say is right is:
a) the lenses would be capable of producing relevant data at this level, which they are not because of the diffraction limit, and b) someone actually looks at data with the same physical dimensions which either implies he's not looking at the picture
Megapixels only take you so far (Score:2)
Re: (Score:3)
Indeed. My fifteen year old Canon 10D with a $50 50mm prime lens will take better photos than pretty much any cell phone camera available today. That said, I'm not likely to take the Canon to the pub, so there's a role for both.
Re: Megapixels only take you so far (Score:2)
The main thing the sensor size gets you is better separation with depth of field. That gives far more artistic options.
If you're shooting something where you want a really large dof, this tiny phone camera sensor will work great.
Putting a sensor like this in a mirrorless, like a m43 mount, might lead to some really nice cameras outside the phone camera realm, too.
Re:Megapixels only take you so far (Score:5, Interesting)
That $7000 Nikon D5 is the see in the dark Batman camera. It's meant for photojournalists who have to shoot in often horrible conditions. Their other camera is a $5000 super resolution 45 MP that generates images with nearly 15 stops of dynamic range and in 14 bit. Both of these cameras produce images that are so far beyond what phones can do. The cameras have their sensors and Analogue to Digital converters fine tuned for their different jobs.
A few years ago one of the Chicago papers laid off all its photographers and gave their reporters iPhones. When the hockey team won the Stanley Cup guess whose front page was an embarrassment.
The sheer amount of light gathering full sized cameras can do justifies their existence.
Re: (Score:2)
That $7000 Nikon D5 is the see in the dark Batman camera.
Or a fast shutter, motion capturing camera for sports and nature photography.
Re: (Score:2)
Oh, the speed! Honestly nicer lenses and better sensor and all that is below my top two reasons for having a DSLR, which are an indirect flash and the speed (meaning shutter lag and burst, not f stops).
How about, no, Sony. (Score:4, Interesting)
Re: How about, no, Sony. (Score:2)
Sony has pioneered this with their a7s line and have some of the best low light sensors made. Phone cameras kind of have a requirement for cramming in pixels due to the size. I'd be happy with a 5mpx camera that's very sensitive to low light, but most people wouldn't.
Re: (Score:2)
Re: (Score:1)
On the other hand I have had 24"x36" prints made from some of my images and there even my 24MP high end DSLR wasn't enough to get the needed resolution.
24MP @ 24"x36" is plenty good unless you're standing, like, 2 ft away. Back up, motherfucker!
Re: (Score:2)
The number of megapixels is only very lightly related to the amount of light that comes in. You're still getting the same light hitting the sensor, and with modern sensor designs very little is wasted on gaps between photosensitive areas ... unlike say 15 years ago. You want to take a half-way decent night shot? Cram as many megapixels as you want in. The noise profile is random across pixels so noise reduction algorithms work true wonders when you have a very fine grain to remove, especially when that nois
Sony XZ(2) with IMX300 sensor... (Score:2)
23 megapixel, 960 fps (IIRC at 720p). Good low light performance for tiny glass (f 2.0).
Already the best mobile phone camera. Higher resolution will just get you clear pixels of blur.
Sony has a long history of keeping the best CCD chips for their own cameras, selling the high defect ones to other camera companies.
Re: (Score:2)
Any evidence for this? There's a big difference between keeping a design exclusive and selling inferior or faulty goods. I can't imagine for a second that the likes of Nikon would accept high defect CCDs.
Re: (Score:2)
They've been notorious for it for 30+ years now.
Sony makes the best CCDs, what is Nikon going to do about it? Sony will likely sell them the next best batch if they pay a little extra. Gives them an edge on Olympus.
Re: Sony XZ(2) with IMX300 sensor... (Score:2)
It's a CMOS, but their 42mpx full frame sensor was developed by Sony and Nikon and both use it in different cameras. What's Nikon going to do? Nikon is going to join them.
Or, not and. (Score:1)
On paper, that means Sony has managed to produce a sensor that combines a huge amount of detail with excellent light capture and low noise levels as well.
No, it doesn't combine. It gives you a choice between a huge amount of detail or excellent light capture or low noise levels. There are only so many photons hitting a given area at a given exposure no matter how intelligently you subdivide the area.
Why do we still believe in magic in the age of technology?
.8 microns? (Score:2)
I have had a full-frame camera with 7 micron pixels for nearly 10 years, and I am still amazed at the pictures clarity it yields. I doubt that .8 micron pixels are ever going to give a good result
Re: (Score:2)
https://www.dxomark.com/huawei... [dxomark.com]
I think the idea is you have more noise, but the noise is smaller than the details you're aiming to capture.
Then you average several pixels and make the best-scoring phone camera there is. The P20 pro only puts out a 10MP image... I bet this one will be used such that it puts out 12MP. Totally reasonable.
Sam
a tiny phone lens can only admit so much light (Score:2)
Cameras in phones have pretty much killed off the idea of a separate "point and shoot" camera for some time now, and yes, phone cameras can do some absolutely amazing stuff, especially combined with things like in-camera HDR, editing on the phone, instant cloud backup, and ability to share your photos pretty much instantly anywhere you have a data signal. That's pretty neat stuff, and as they say "the camera you use is the one that you have with you". So they are a great blending of two devices in one, and
Re: (Score:2)
A micro 4/3 is a nice, functional, smaller choice.
Often the same sensor as the small sensor 35mm DSLR, which becomes full field in the smaller form factor.
But, of course it means all new lenses. I had real good luck with used from japan vendors on Amazon. Some of those guys just NEED the newest stuff. Year old, half price. Prime lens, F0.7 for about $200.
You still have to carry it, about half the bulk.
Re: (Score:2)
Re: (Score:2)
Because consumers don't understand (Score:5, Interesting)
That said Sony does make some damn fine sensors but no one who knows about optics and sensors really expects this to compete with even entry level DSLRs or mirror-less interchangeable lens cameras, let alone those monster digital medium formats from Hasselblad or Pentax. Instead it will be something for consumers to get into a phone pissing contest over and believe that they can take pictures just as good as a pro can.
Re: (Score:2)
Thank you. I wanted to say something about diffraction but you are already there. I wonder what actual resolution can be achieved with a cell phone optic. I don't see the sense of adding pixels beyond that.
Yeah, pixel war... (Score:5, Insightful)
Then came in the electrics, the puny wheezy golf cart electrics.... Beats them hollow in their own game. An electric SUV beats an Alpha Romeo Spyder, while towing an Alpha Romeo Spyder!
Wish someone will make a CCD with a dynamic range 3 orders of mag better than the crappy ones we have.
Re: (Score:2)
Wish someone will make a CCD with a dynamic range 3 orders of mag better than the crappy ones we have.
What for? As it stands currently the biggest problem with have with the images our existing CCDs are recording is processing them in a way to display that detail. We already have the ability to shoot into the sun while seeing into the shadows. What will even more dynamic range show? That glass is imperfect and flaring limits our ability to see?
Zoom? (Score:3)
While I know this high pixel count is worthless, can anyways with actual knowledge of the technology tell me if this would help with digital zoom? If my cellphone camera could offer me a 10x zoom producing a 4 Megapixel image, this is something I might consider buying.
Re: (Score:2)
While I know this high pixel count is worthless
Why do you know that? Higher pixel counts do wonders for the ability to process data. Control of the individual pixels even more so. HDR, noise reduction, better colour reproduction through different and very lossy interpolation across the beyer sensor, software based vibration reduction all benefit from higher pixel counts.
But ultimately the answer is a: not really. You can cram as many pixels as you want into your sensor. That 3mm tiny camera lens will limit your sharpness and your practical ability to di
Quad Bayer (Score:2)
It's a trick (Score:2)
Even if it's a Sony.
Re: (Score:1)
Won't believe until I see it (Score:3)
There is a diffraction limit which sets in when the pixel size is smaller than 1.22*wavelength*f_ratio. The wavelength range is 0.4-0.7 micron (let us take middle wavelength of 0.55 micron). Best f_ratio that I have seen is 1.8, so the limit kicks in at at around 1.2 micron pixel size. The limit is for far off objects and you get little better when nearby assuming you have a perfect lens.
In practice, it will be impossible to beat 16 MP SLR camera in terms of resolution. This is why mega zoom camera cannot give as good resolution of moon as some of the SLR with large lens can give.
Re: (Score:2)
Diffraction limit is only related to image sharpness. There's a lot more to having more megapixels than sharpness when you start exploring the oportunities of signal processing.
quality and MP (Score:1)
Lumia 1020 (Score:2)
The "camera" smart phone to compare against isn't one of the new ones, but the Nokia Lumia 1020. It actually had a good camera: good lens, 41 megapixel sensor, good light/shadow sensitivity, real flash, good camera software. Downside was it was a Window's Phone (anyone who actually used a Windows Phone probably liked the operating system and hated the lack of a good app store), and it was oddly shaped. My 1020 stopped working as a phone a while ago, but it's still the best camera I own.
Pixel Count? (Score:2)
How does this help when performance is limited by the lens? My 15 megapixel camera has more coma and chromatic aberration than the sensor will support.