Forgot Password
Pentax Camera Forums Home
 

Reply
Show Printable Version Search this Thread
02-28-2012, 04:17 AM   #76
Pentaxian
gazonk's Avatar

Join Date: Feb 2009
Location: Oslo area, Norway
Photos: Gallery | Albums
Posts: 3,746
QuoteOriginally posted by Erik Quote
It is ALSO a marketing lie when they claim that the pixel binning process somehow turns individial pixels in the resampled output picture into "true RGB pixels".
I think it's really a stretch to call that a lie. If each pixel contains the information from both red, blue and green photo sites, does it really matter if those photo sites are stacked on top of each other (like in a Foveon sensor) or beside each other?

(also, it's not clear from the white paper how the oversampling is done - it's not obvious that it's simply done by demosaicing and then pixel binning)

02-28-2012, 04:25 AM   #77
Veteran Member




Join Date: Nov 2008
Location: Umeċ, Sweden
Posts: 755
QuoteOriginally posted by gazonk Quote
I think it's really a stretch to call that a lie. If each pixel contains the information from both red, blue and green photo sites, does it really matter if those photo sites are stacked on top of each other (like in a Foveon sensor) or beside each other?
OK -- but in that case, where's the cutoff point? As you "zoom in" on this camera, you go from binning, say, 16 pixels (actually, 16 groups of 4 monochrome photosites) to 1 to eventually 4 to 1, and at last, 1 to 1.* Where exactly do the "true RGB" pixels end? I'm just advocating calling a spade a spade, because to me, at no point are those actually "true RGB" pixels. They're a close approximation at best. So yes, it's a lie, because it's not actually true.


* These numbers are grabbed out of thin air just to demonstrate the point; I don't feel like actually doing the math


QuoteOriginally posted by gazonk Quote
(also, it's not clear from the white paper how the oversampling is done - it's not obvious that it's simply done by demosaicing and then pixel binning)
It would probably be done before demosaicing, I'd hope! So with that said, my previous assertion that downsampling in Photoshop is pretty much the same thing was a bit over-simplified.

Last edited by Erik; 02-28-2012 at 04:38 AM.
02-28-2012, 04:47 AM   #78
Veteran Member




Join Date: Nov 2008
Location: Umeċ, Sweden
Posts: 755
And also, just to set the record straight -- if I sound a bit negative in this thread, it's only because I'm allergic to marketing-speak and wish to demystify some things about the technology involved. The camera module still looks pretty profoundly awesome compared to every compact digital camera and cameraphone ever. Let it be known that if I ever buy a smartphone, this is a very likely contender. I would LOVE to have decent imaging capabilities always in my pocket.

Right now, Nikon's CX system is looking pretty stupid.

Check out this 5MP ISO 800 picture. That is some amazing, amazing performance. I hate to drag out the old "the noise profile looks like film grain!!!" thing, but... the noise profile looks like film grain!!!
02-28-2012, 05:13 AM   #79
Veteran Member
Jodokast96's Avatar

Join Date: Oct 2007
Location: Erial, NJ USA
Photos: Gallery | Albums
Posts: 2,134
QuoteOriginally posted by séamuis Quote
this is clearly an incredible camera, but I don't think it will do anything to help Nokia, because people just don't buy smartphones for the camera. the OS is the biggest deciding factor and this is precisely where the Nokia and all current Nokias fall short. not that symbian is bad, though I haven't used Symbian 'Belle'. all of the latest Nokias have been exceptional, nay, outstanding phones with the best cameras in the industry and beyond. but who besides current nokia users is likely to go buy the 808?
As someone else pointed out to me, investors agree with you.

https://www.google.com/finance?client=ob&q=NYSE:NOK

02-28-2012, 05:18 AM   #80
Veteran Member
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,871
Original Poster
QuoteOriginally posted by Erik Quote
But the fact is that the "equivalent" focal length never ever changes.
False
-> Falk Lumo: Camera equivalence

Note: I won't discuss it any further here in this context. All I have to say is in the link above.
02-28-2012, 05:25 AM   #81
Veteran Member
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,871
Original Poster
QuoteOriginally posted by Erik Quote
if I sound a bit negative in this thread, it's only because I'm allergic to marketing-speak and wish to demystify some things about the technology involved.
I fully understand this emotion. However, in this paricular case, I disagree.

Nokia will fail with their approach if potential customers fail to see the point. And let's face it, 99% of customers don't have any demand beyond 8MP. Let alone in a phone.

So, their speech must be that their technology does indeed replace the optical zoom which phone cameras lack.

IMHO, they could have been a bit more aggressive to highlight the unprecedented (for a P&S) low light capabilities and image quality at the wide end. To say that the blurring effect of the Bayer filter becomes invisible (at the wide part) is important and correct.

And they move on to explain in ample detail how this is done and it is binning(*) and in-cropping actually.
(*) I am not sure actually that they do in-sensor binning. It more looks like traditional in-camera downsampling which is a good thing.

I think this is one of the rare cases where marketing speech fulfils a useful function. To make people understand what it is good for.


Btw, 4x downsampling with a Bayer sensor does indeed deliver the same full color RGB pixel quality a Foveon sensor does. With the added advantage that it delivers 4x higher resolution in the luminosity channel.

Last edited by falconeye; 02-28-2012 at 05:38 AM.
02-28-2012, 05:41 AM   #82
Pentaxian
gazonk's Avatar

Join Date: Feb 2009
Location: Oslo area, Norway
Photos: Gallery | Albums
Posts: 3,746
QuoteOriginally posted by falconeye Quote
Btw, 4x downsampling with a Bayer sensor does indeed deliver the same full color RGB pixel quality a Foveon sensor does. With the added advantage that it delivers 4x higher resolution in the luminosity channel.
The announcement of this sensor has got me wondering if a 48-64 mp APS-C sensor actually could make a whole lot of sense. It could be used at full resolution at low ISOs and with 4x binning at higher ISOs, giving 12-16 mp. For shooting at low ISOs, it could make a wide angle prime (say, in the 16-24mm range, depending on how wide you need) with very good center sharpness the ideal walk-around lens because of the cropping possibilities.

02-28-2012, 05:53 AM   #83
Veteran Member
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,871
Original Poster
QuoteOriginally posted by Unsinkable II Quote
At full resolution in good light, this thing will likely take incredibly detailed landscapes and cityscapes.
Yes and no.

At full resolution, the 1.4µm pixels are rather close to the diffraction limit (which is at 0.92µm). Anything closer to ~1.8µm will be visibly affected by diffraction. You will have to sharpen assuming there are no other sources of blur. The good side is that this replaces the Bayer AA filter and avoids Moiré artefacts. And should be good enough for 16MP in good quality at the pixel level.

It would then depend on the quality of this Zeiss lens in the corners. I am pretty sure it outresolves the sensor in the center (as this is mandatory to make the zoom feature work). But in the wide angle corners, it may have been specified for a 5MP resolution.

So, assuming it represents an ultra compact 16MP K-5 with 17mm prime (although with at least ISO 400 as its lowest ISO setting), this statement relies on lenses.

Now, I've seen 808 samples which are indeed soft in the corners. However, the same holds true for DA15 samples in the corners.

So, overall image quality may indeed be comparable to an ISO 400 DA15 K-5.
02-28-2012, 06:16 AM   #84
Veteran Member
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,871
Original Poster
QuoteOriginally posted by gazonk Quote
For shooting at low ISOs, it could make a wide angle prime (say, in the 16-24mm range, depending on how wide you need) with very good center sharpness
Theoretically, yes.

But this 808 lens is a 26-80mm F9-28 equivalent (35mm) diffraction-limited (center) lens (assuming cropping and binning to a constant size image). The high F-stop at the tele end shows you how diffraction-limited tha approach is. This is why Nokia claims 5MP only and that already implies much reduced pixel contrast at the tele end then (although still sharpenable). In the end, you see that the approach works well for a 5MP target size.

However, if you want, e.g. sharp 20MP output, you have much higher requirements. E.g., you at least would need a lens which is a 26-80mm F4.5-14 equivalent (35mm) and diffraction limited. This is doable with full frame. Well, a 160MP "D800" with 24-70/2.8 stopped down to f/4 (or the 24/1.4 stopped down to f/4) is exactly this (yes, these lenses would resolve a 160MP sensor!). However, I heavily doubt that the DA*16-50/2.8 wide open (or any 16mm prime at F2.8) would outresolve a 160MP sensor in the center, the analog requirement. Even if the lens does, the focus accuracy with phase detect won't. Btw, "D800" would need contrast AF too, using the high resolution sensor capabilities 160MP would offer.

So, you can do it with crop 3 for 5MP, or crop 1 for 20MP, but with crop 1.5 it will be impossible to obtain the current image resolutions (16MP). FourThird (crop 2) with contrast AF may be able to deliver 10MP using this method with 80MP sensors.

Last edited by falconeye; 02-28-2012 at 06:24 AM.
02-28-2012, 06:18 AM   #85
Veteran Member




Join Date: Nov 2008
Location: Umeċ, Sweden
Posts: 755
QuoteOriginally posted by falconeye Quote
Since you will not discuss this further, I will end with this then: false, and your link is not relevant to what I am saying.
02-28-2012, 06:35 AM   #86
Veteran Member




Join Date: Nov 2008
Location: Umeċ, Sweden
Posts: 755
For those who are still curious, though, it is trivial to demonstrate that any attempt to put this "zoom by cropping" technology into terms of focal length is not useful or correct:



Assuming the camera is at the same distance from the subject in all three pictures, the facts are: the focal length never changes, the perspective never changes, and the depth of field never changes. The last two are defining properties of a given focal length. Therefore, it is deceptive to talk about "equivalent focal length" ever differing, regardless of "zoom level", in this design. It simply does not. The sensor always sees the exact same image, you are just selecting which portion you want to crop.
02-28-2012, 07:02 AM   #87
Pentaxian
gazonk's Avatar

Join Date: Feb 2009
Location: Oslo area, Norway
Photos: Gallery | Albums
Posts: 3,746
QuoteOriginally posted by Erik Quote
Assuming the camera is at the same distance from the subject in all three pictures, the facts are: the focal length never changes, the perspective never changes, and the depth of field never changes.
It's NOT true that DoF doesn't change. The more you crop an image, the shallower the DoF becomes because of the enlargement.

It IS true that perspective doesn't change. But the same is true if you use different lenses. Assuming a rectilinear lens, the perspective ONLY depends on the distance to the subject - it's purely a matter of geometry, not lens focal length! (I can't believe how many times this has to be said!)
02-28-2012, 07:28 AM   #88
Veteran Member




Join Date: Nov 2008
Location: Umeċ, Sweden
Posts: 755
QuoteOriginally posted by gazonk Quote
It's NOT true that DoF doesn't change. The more you crop an image, the shallower the DoF becomes because of the enlargement.
OK. I understand this is technically true, but I cannot agree that it is a useful way to think. If I have a photo taken at f/2.8 and crop it severely, it would still be wrong to me to say "this is picture with the depth of field of a f/0.1 lens". It's just a crop. I will concede that I was technically wrong to say depth of field does not change, though.

QuoteOriginally posted by gazonk Quote
It IS true that perspective doesn't change. But the same is true if you use different lenses. Assuming a rectilinear lens, the perspective ONLY depends on the distance to the subject - it's purely a matter of geometry, not lens focal length! (I can't believe how many times this has to be said!)
I understand that, but this still means that if this sensor was cropped to what you might call "100 mm-e", you still often cannot get pictures with the same perspective as you could with an actual 100 mm-e lens -- so talking about a camera that functions by cropping in terms of "equivalent focal lengths" is misleading.

That the focal length never changes here is fact, so why muddy the waters with "equivalents", since it in no way acts like an actual zoom lens?
02-28-2012, 07:46 AM - 1 Like   #89
Veteran Member
bobmaxja's Avatar

Join Date: Mar 2010
Location: Laval, Quebec Canada
Posts: 2,171
Hey guy , you can fight all you want but the buyer of these phone dont care. They will not do comparaison with any DSLR, or DoF ,etc.
Their friend will say , Wow , my phone can not do this
02-28-2012, 07:55 AM   #90
Veteran Member
froeschle's Avatar

Join Date: Sep 2007
Location: Germany
Photos: Gallery
Posts: 552
QuoteQuote:
[...] so why muddy the waters with "equivalents", since it in no way acts like an actual zoom lens? [...]
The idea of Nokia was to replace "a decent zoom lens in a camera phone" by "zooming by very high resolution image sensor". Thus, the obvious question is, which system this mobile could replace. Or in other words, which camera would give identical results, i.e. equivalent photos (see falconeye and e.g. Equivalence )

Picture of sensor and lens: http://allthingsd.com/files/2012/02/808-with-808.png
From Nokia's 41-Megapixel - 5 years in the making
Reply

Bookmarks
  • Submit Thread to Facebook Facebook
  • Submit Thread to Twitter Twitter
  • Submit Thread to Digg Digg
Tags - Make this thread easier to find by adding keywords to it!
camera, nokia, pureview, segment, sensor
Thread Tools Search this Thread
Search this Thread:

Advanced Search


Similar Threads
Thread Thread Starter Forum Replies Last Post
Hasselblad's 200 megapixel H4D-200MS camera now shipping jogiba Canon, Nikon, Sony, and Other Camera Brands 3 07-01-2011 09:23 AM
The benefit of a SmartPhone for a photography business dugrant153 General Talk 6 05-13-2010 11:26 PM
Meet the 570-megapixel digital camera joodiespost General Talk 3 01-15-2010 12:10 AM
[WTF] Microsoft`s 196 Megapixel Camera Reportage General Talk 2 04-12-2009 08:33 AM
Pentax's 30 megapixel digital camera alderfall Pentax News and Rumors 31 03-27-2009 12:40 PM



All times are GMT -7. The time now is 09:47 PM. | See also: NikonForums.com, CanonForums.com part of our network of photo forums!
  • Red (Default)
  • Green
  • Gray
  • Dark
  • Dark Yellow
  • Dark Blue
  • Old Red
  • Old Green
  • Old Gray
  • Dial-Up Style
Hello! It's great to see you back on the forum! Have you considered joining the community?
register
Creating a FREE ACCOUNT takes under a minute, removes ads, and lets you post! [Dismiss]
Top