Ad
Ad
Ad
Pages: « 1 [2] 3 4 ... 8 »   Bottom of Page
Print
Author Topic: What about 36MP DSLRs?  (Read 29587 times)
hjulenissen
Sr. Member
****
Offline Offline

Posts: 1675


« Reply #20 on: December 21, 2011, 07:14:51 AM »
ReplyReply

Hi,

It depends. Shot noise is essentially independent of pixel size for a given size of sensor and Quantum Efficiency. Think of shot noise as noise lights and mid tones.

Regarding read noise (darkest part of the image) the math is different. Increasing pixel size will reduce noise more than just resampling to lower density.
Downsampling works well as an abstraction when discussing these things, but I dont think that is how one should reduce noise. Rather, one should use a good noise reduction alogrithm.

A good noise reduction algorithm could expolit human vision and camera sensor deficiencies in a better, more complex manner than simply increasing the sensel size (or downsampling). The comparision of large sensels vs small sensels + NR is difficult.

-h
Logged
torger
Sr. Member
****
Offline Offline

Posts: 1398


« Reply #21 on: December 21, 2011, 07:36:46 AM »
ReplyReply

My experience is that in practice noise is not the first thing that gives you practical problems when making prints, but instead poor color reproduction is, you get bleaker and bleaker colors and poor tonal range.

Increase of noise levels and reduction of tonal range follow hand in hand though so to compare two systems it is typically enough to look at noise. However if you apply noise reduction algorithms in post it seems to me that you don't just kill some fine detail, you kill color too. Therefore I'm very skeptical about noise reduction in post.

Hardware noise reduction can be good however, when information about the the local noise profile makes it possible to reduce noise without much loss in real signal information.
Logged
hjulenissen
Sr. Member
****
Offline Offline

Posts: 1675


« Reply #22 on: December 21, 2011, 06:00:19 PM »
ReplyReply

...However if you apply noise reduction algorithms in post it seems to me that you don't just kill some fine detail, you kill color too. Therefore I'm very skeptical about noise reduction in post.

Hardware noise reduction can be good however, when information about the the local noise profile makes it possible to reduce noise without much loss in real signal information.
I was thinking about the post below when writing my response. I believe that any sensible NR should be better than downsampling when trying to reduce noise, and that downsampling for noise reduction is not a good practice. But it makes for a very reasonable "discussion tool" for comparing different camera-choice/processing trade-offs.
Would down sampling in post processing from 36mp to say 24 mp improve the noise? i.e. would it be possible to shoot at a higher ISO for the same noise? If this is true how does one calculate by how much i.e. image shot at ISO 3200/36mp would be similar in noise to 24mp/??mp.
See also these (heated) discussions:
dpreview - downsampling reduces noice! - uh noise! - con't.
dpreview - downsampling reduces noice!
« Last Edit: December 21, 2011, 06:03:12 PM by hjulenissen » Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3553


« Reply #23 on: December 21, 2011, 06:45:33 PM »
ReplyReply

I was thinking about the post below when writing my response. I believe that any sensible NR should be better than downsampling when trying to reduce noise, and that downsampling for noise reduction is not a good practice.

Hi,

I agree. If (!) downsampling is required and it results in reducing the noise, that's fine. However I don't view it as a means of noise reduction because it also loses a lot of other information. What is important to understand is that noise has a power spectrum along the spatial frequency dimension. So noise can vary at the various spatial frequencies, and downsampling changes the spectrum.

Quote
But it makes for a very reasonable "discussion tool" for comparing different camera-choice/processing trade-offs.

Well, fair comparisons require the elimination of variables that are specific to the items under investigation. If output size is one of those variables, then it needs to be eliminated (although one also changes variables that are not included in the specific comparison, such as output resolution).

Cheers,
Bart
Logged
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7308


WWW
« Reply #24 on: December 21, 2011, 07:05:47 PM »
ReplyReply

Hi,

In my view the question is what advantage/disadvantage smaller pixels have versus larger pixels. If we look at the two images at the same size or print at the same size either or both of the images will be resized. So we don't discuss downsizing as a method of reducing noise per se but to discuss how noise would be affected if we don't compare both images at actual pixels.

Best regards
Erik


Hi,

I agree. If (!) downsampling is required and it results in reducing the noise, that's fine. However I don't view it as a means of noise reduction because it also loses a lot of other information. What is important to understand is that noise has a power spectrum along the spatial frequency dimension. So noise can vary at the various spatial frequencies, and downsampling changes the spectrum.

Well, fair comparisons require the elimination of variables that are specific to the items under investigation. If output size is one of those variables, then it needs to be eliminated (although one also changes variables that are not included in the specific comparison, such as output resolution).

Cheers,
Bart
Logged

BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3553


« Reply #25 on: December 21, 2011, 07:34:26 PM »
ReplyReply

Hi,

In my view the question is what advantage/disadvantage smaller pixels have versus larger pixels. If we look at the two images at the same size or print at the same size either or both of the images will be resized. So we don't discuss downsizing as a method of reducing noise per se but to discuss how noise would be affected if we don't compare both images at actual pixels.

Hi Erik,

Fair enough, but I have difficulty understanding why one gets involved with the drawbacks of smaller sensels (increased per-pixel noise, and larger file size) unless physically larger output with higher resolution is the (optional) goal, or as a means to reduce aliasing.

There are better methods for potentially reducing noise than downsampling.

Cheers,
Bart
Logged
hjulenissen
Sr. Member
****
Offline Offline

Posts: 1675


« Reply #26 on: December 21, 2011, 07:35:24 PM »
ReplyReply

Hi,

In my view the question is what advantage/disadvantage smaller pixels have versus larger pixels. If we look at the two images at the same size or print at the same size either or both of the images will be resized. So we don't discuss downsizing as a method of reducing noise per se but to discuss how noise would be affected if we don't compare both images at actual pixels.

Best regards
Erik


I think that a question that many seems to be interested in is "if I use camera A to make this image, featuring 24MP and APS-C sensor, or camera B, featuring 12MP and APS-C sensor, which will look best printed as A2, everything else equal?" Many have opinions about the results. Some because of long practical, relevant experience. Some because of long practical, irrelevant experience. Some because of solid theoretical understanding. Some because of faulty theoretical understanding.

For making fair side-by-side images that people will tend to scrutinize at 1:1 pixel, it makes sense to me 1) to downsample the hirez one (to consider low-light relative performance at smallish printsize vs original lorez image) and 2)to upsample the lorez one (to consider level of details at larger print sizes vs hirez one). One might say that this comparision is unfair: you should rather resample directly to whatever the viewing resolution is, and apply optimal NR to both. But what is optimal NR and how is it defined?


BTW, as I said at dpreview, I think that downsampling can be viewed as similar to noise reduction. What it does it similar to noise reduction (reduce noise and detail), only that it does not do it particulary well, and was never meant to by the designer. The "noise reducing" capabilities of downsampling is limited to the inherent prefilter, so when studying downsamplings effects on image noise it might be wise to consider only the (more or less) brickwall lowpassfilter. An favored ad-hoc choice seems to be the 2-3rd order lanczos response (sinc windowed by a sinc).

-h
Logged
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7308


WWW
« Reply #27 on: December 21, 2011, 10:46:59 PM »
ReplyReply

Hi,

If we assume that printer resolution is 360 PPI and we print at A2 the pixel dimension for the print will be around 8300x5900. That is about 50 MPixels. So for A2 prints pretty much everything this side of P65+ will be "upprezzed" in print. The resizing would either be done in LR/PS or by the printer driver.


The above calculation is based 42x59 cm print size, in many cases the actual image would be significantly smaller.

Best regards
Erik

I think that a question that many seems to be interested in is "if I use camera A to make this image, featuring 24MP and APS-C sensor, or camera B, featuring 12MP and APS-C sensor, which will look best printed as A2, everything else equal?" Many have opinions about the results. Some because of long practical, relevant experience. Some because of long practical, irrelevant experience. Some because of solid theoretical understanding. Some because of faulty theoretical understanding.

For making fair side-by-side images that people will tend to scrutinize at 1:1 pixel, it makes sense to me 1) to downsample the hirez one (to consider low-light relative performance at smallish printsize vs original lorez image) and 2)to upsample the lorez one (to consider level of details at larger print sizes vs hirez one). One might say that this comparision is unfair: you should rather resample directly to whatever the viewing resolution is, and apply optimal NR to both. But what is optimal NR and how is it defined?


BTW, as I said at dpreview, I think that downsampling can be viewed as similar to noise reduction. What it does it similar to noise reduction (reduce noise and detail), only that it does not do it particulary well, and was never meant to by the designer. The "noise reducing" capabilities of downsampling is limited to the inherent prefilter, so when studying downsamplings effects on image noise it might be wise to consider only the (more or less) brickwall lowpassfilter. An favored ad-hoc choice seems to be the 2-3rd order lanczos response (sinc windowed by a sinc).

-h
Logged

fotometria gr
Sr. Member
****
Offline Offline

Posts: 568


WWW
« Reply #28 on: December 22, 2011, 05:40:51 PM »
ReplyReply

Hi,

Here is an interesting posting by Diglloyd: http://www.diglloyd.com/blog/2011/20111121_1-Megapixels.html

Small comment. Going from 24 MP to 36 MP is a minor step. It's a 22% increase in resolution. No dramatic effects expected.

Best regards
Erik
Going from 16mpx to 24 is the same minor step.... Cheers, Theodoros. www.fotometria.gr
Logged
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7308


WWW
« Reply #29 on: December 22, 2011, 10:27:28 PM »
ReplyReply

Hi,

Yes, indeed! We are climbing the ladder in small steps. Going from 16MP to 24MP is a small step, and so going from 24 MP to 36MP. Going from 16 to 36 MP is probably a significant step. The observation I made in my "demo" was that the advantage of 36 MP over 24 MP was clearly visible on charts, but far less on the red flower include in the picture. Someone observed that the greens were better on the "36MP" image. This may possibly depend on diffraction. Red light will show much more diffraction than green light. Or it could depend on chromatic aberration.

Making pixels smaller allows for weaker OLP-filtering. I have seen some Moiré on my 16 MP APS-C camera and some Moiré is also present in Michael Reichmans test images with the NEX-7, 24MP APS-C.

Best regards
Erik


Going from 16mpx to 24 is the same minor step.... Cheers, Theodoros. www.fotometria.gr
Logged

fotometria gr
Sr. Member
****
Offline Offline

Posts: 568


WWW
« Reply #30 on: December 23, 2011, 01:55:55 PM »
ReplyReply

Hi,

Yes, indeed! We are climbing the ladder in small steps. Going from 16MP to 24MP is a small step, and so going from 24 MP to 36MP. Going from 16 to 36 MP is probably a significant step. The observation I made in my "demo" was that the advantage of 36 MP over 24 MP was clearly visible on charts, but far less on the red flower include in the picture. Someone observed that the greens were better on the "36MP" image. This may possibly depend on diffraction. Red light will show much more diffraction than green light. Or it could depend on chromatic aberration.

Making pixels smaller allows for weaker OLP-filtering. I have seen some Moiré on my 16 MP APS-C camera and some Moiré is also present in Michael Reichmans test images with the NEX-7, 24MP APS-C.

Best regards
Erik


Hence, there is no problem with resolution but rather a problem of "pixel definition", where bigger pixels have the advantage in DR and noise. Of course smaller pixels do behave better as far as moire is concerned but is this enough to ignore DR and noise? Shouldn't the manufacturers look for different ways than oversizing sensors to overcome moire? I remember you believe that a 4μm sensor can do both but I think that looking at what sensor manufactures do (with the recent example of 1DX or MF manufactures) they (and I) think different. I also think that its irrelevant to judge from an aps-c sensor, since there you don't have the huge angle that is required on a full frame sensor for light to enter the pixel well at the edge of the frame. Regards, Theodoros. www.fotometria.gr
 P.S. Clearly I don't believe there will be any 36mpx D800, I don't believe they will change a winning horse/policy, its stupid marketing to do so. There may be such a sensor as an alternative to the 16mpx one on the D4, which I am almost sure that will have interchangeable sensors, but not on a D700 replacement, the 16mpx sensor will be more than enough there.
Logged
BernardLanguillier
Sr. Member
****
Offline Offline

Posts: 7855



WWW
« Reply #31 on: December 23, 2011, 06:15:40 PM »
ReplyReply

Hence, there is no problem with resolution but rather a problem of "pixel definition", where bigger pixels have the advantage in DR and noise. Of course smaller pixels do behave better as far as moire is concerned but is this enough to ignore DR and noise? Shouldn't the manufacturers look for different ways than oversizing sensors to overcome moire? I remember you believe that a 4μm sensor can do both but I think that looking at what sensor manufactures do (with the recent example of 1DX or MF manufactures) they (and I) think different. I also think that its irrelevant to judge from an aps-c sensor, since there you don't have the huge angle that is required on a full frame sensor for light to enter the pixel well at the edge of the frame. Regards, Theodoros. www.fotometria.gr
 P.S. Clearly I don't believe there will be any 36mpx D800, I don't believe they will change a winning horse/policy, its stupid marketing to do so. There may be such a sensor as an alternative to the 16mpx one on the D4, which I am almost sure that will have interchangeable sensors, but not on a D700 replacement, the 16mpx sensor will be more than enough there.

Technology has been proven to have a larger impact on DR than theoretical physics, has it not?

The gap of DR between a D3x and a 1Ds MkI is much larger than the gap between MF and 35mm ever was in a given generation of sensors.

Cheers,
Bernard

« Last Edit: December 23, 2011, 06:17:43 PM by BernardLanguillier » Logged

A few images online here!
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7308


WWW
« Reply #32 on: December 23, 2011, 09:23:33 PM »
ReplyReply

Hi,

I have never discussed downsizing the image as method of noise reduction. What I say is that shot noise mainly depends on sensor size and not pixel size, when the image is printed. I also assume that we are interested in printing large. The original posting on this thread refers to an experiment I made:

http://echophoto.dnsalias.net/ekr/index.php/photoarticles/60-what-about-36mp-dslrs

The main advantages with smaller pixels I see is mainly:

- Aliasing effects are reduced
- OLP filtering is effectively reduced
- The image will need less sharpening and also respond better to sharpening, at least at low ISO

But I'm pretty sure that OLP-filtering is still needed. Aliasing would only be possible if the lens transfers significant MTF at pixel pitch and I'm surprised how well lenses hold up to modern sensors. On the NEX-7 "BJanes" measured 109 lp/mm on the DPreview comparison image. I have seen some artifacts and also color Moiré on my Sony Alpha 55 SLT. Sometimes it shows up as detail that is probably unnatural.

There has been some discussion on negative side effects of smaller pixels, with possible crosstalk between cells.

This article may be of interest regarding optimal pixel size: http://isl.stanford.edu/~abbas/group/papers_and_pub/pixelsize.pdf

Anyway the issue is complex. One aim is the best possible reproduction of detail, but that also needs careful work. Camera on tripod, MLU, exact focusing and using moderate apertures to avoid diffraction. The other extreme is low light shooting under free hand condition. Sensor FWC will be underutilized, and read noise much more visible.

It seems that the cameras of today having best high ISO performance have relatively large pixels, but don't have very good DR at low ISO, because of high amount of readout noise. It is also my understanding that high ISO cameras normally use pre amplifiers to achieve medium ISO values, but that means that the signal coming out of the sensor must be cleaner than the unamplified signal going into the ADC, otherwise amplification would not help. Why then is no preamplification used at base ISO?

Best regards
Erik

Hi Erik,

Fair enough, but I have difficulty understanding why one gets involved with the drawbacks of smaller sensels (increased per-pixel noise, and larger file size) unless physically larger output with higher resolution is the (optional) goal, or as a means to reduce aliasing.

There are better methods for potentially reducing noise than downsampling.

Cheers,
Bart
« Last Edit: December 24, 2011, 03:34:01 AM by ErikKaffehr » Logged

ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7308


WWW
« Reply #33 on: December 24, 2011, 03:38:47 AM »
ReplyReply

Hi,

I don't think technology trumps physics, except for placebo effects. Better technology allows better utilization of limitation set by physics, and technology would not be possible without the underlaying physics.

Best regards
Erik


Technology has been proven to have a larger impact on DR than theoretical physics, has it not?

Cheers,
Bernard


Logged

fotometria gr
Sr. Member
****
Offline Offline

Posts: 568


WWW
« Reply #34 on: December 24, 2011, 03:55:14 AM »
ReplyReply

Technology has been proven to have a larger impact on DR than theoretical physics, has it not?

The gap of DR between a D3x and a 1Ds MkI is much larger than the gap between MF and 35mm ever was in a given generation of sensors.

Cheers,
Bernard


Up to an extend yes, but this doesn't mean that the engineers should ignore physics, it rather means that they better keep a balance that improves both resolution/moire and DR/noise, thats what they are doing up to now isn't it? OTOH look at Canon, they seem to be changing their up to now policy, obviously because they favored resolution more than they should to the direct competition which has beat them in DR/noise. That's why I am sure that in Nikon they won't change a winning horse. I have used the d3x a lot in my studio, to be honest I whould prefer the A900 for the same task, don't you find that Nikon overdid it a bit at in-camera manipulation to give a more "balanced" camera? The A900 seems to be both a little sharper and with slightly better highlight DR. At the end I decided to stay with my Contax645/Imacon 528c combo for the task, although I am a Nikon user, it was clearly better than both, but on moire. Regards, Theodoros. www.fotometria.gr
Logged
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7308


WWW
« Reply #35 on: December 24, 2011, 04:18:37 AM »
ReplyReply

Hi,

I think this issue with DR is a bit overblown. It is a technical term, essentially measurable maximum possible signal/minimum noise.

Best regards
Erik


Up to an extend yes, but this doesn't mean that the engineers should ignore physics, it rather means that they better keep a balance that improves both resolution/moire and DR/noise, thats what they are doing up to now isn't it? OTOH look at Canon, they seem to be changing their up to now policy, obviously because they favored resolution more than they should to the direct competition which has beat them in DR/noise. That's why I am sure that in Nikon they won't change a winning horse. I have used the d3x a lot in my studio, to be honest I whould prefer the A900 for the same task, don't you find that Nikon overdid it a bit at in-camera manipulation to give a more "balanced" camera? The A900 seems to be both a little sharper and with slightly better highlight DR. At the end I decided to stay with my Contax645/Imacon 528c combo for the task, although I am a Nikon user, it was clearly better than both, but on moire. Regards, Theodoros. www.fotometria.gr
Logged

fotometria gr
Sr. Member
****
Offline Offline

Posts: 568


WWW
« Reply #36 on: December 24, 2011, 04:53:40 AM »
ReplyReply

Hi,

I think this issue with DR is a bit overblown. It is a technical term, essentially measurable maximum possible signal/minimum noise.

Best regards
Erik


I suppose it differs from one photographer to another, to me, highlight DR is the most important from all factors that influence photography, I find it on the best sensors to be at least a stop behind film, that's the reason I still use film in some cases. Even the Fuji s5 which is in a class of it's own among digital on that matter (my favorite among my DSLRs), is behind film. I wish/hope that the new generation of DSLRs will fill the gap! Regards, Theodoros. www.fotometria.gr
Logged
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7308


WWW
« Reply #37 on: December 24, 2011, 05:12:33 AM »
ReplyReply

Hi,

There you are! For you DR is about highlight detail. When you compare with film, do you compare with Slide, BW or print film?

Fuji has a neat trick of extending DR with essentially a dual set of pixels in their sensors, with one set for highlight.


http://www.dxomark.com/index.php/Publications/DxOMark-Insights/Super-CCD

Best regards
Erik


I suppose it differs from one photographer to another, to me, highlight DR is the most important from all factors that influence photography, I find it on the best sensors to be at least a stop behind film, that's the reason I still use film in some cases. Even the Fuji s5 which is in a class of it's own among digital on that matter (my favorite among my DSLRs), is behind film. I wish/hope that the new generation of DSLRs will fill the gap! Regards, Theodoros. www.fotometria.gr
« Last Edit: December 24, 2011, 05:14:49 AM by ErikKaffehr » Logged

fotometria gr
Sr. Member
****
Offline Offline

Posts: 568


WWW
« Reply #38 on: December 24, 2011, 06:38:37 AM »
ReplyReply

Hi,

There you are! For you DR is about highlight detail. When you compare with film, do you compare with Slide, BW or print film?

Fuji has a neat trick of extending DR with essentially a dual set of pixels in their sensors, with one set for highlight.


http://www.dxomark.com/index.php/Publications/DxOMark-Insights/Super-CCD

Best regards
Erik


Negative only, B&W or Color, DR is worst on slides. I wish they would have improved on Fuji's technology further! PHEEWW what a camera this would prove to be with another five years of development! Imagine a 30mpx FF sensor (no need to change the size of either R or S sensors), with further improvement in color, DR and noise! Merry Christmas, Theodoros. www.fotometria.gr
Logged
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7308


WWW
« Reply #39 on: December 24, 2011, 07:09:16 AM »
ReplyReply

Hi,

I don't know why Fuji didn't develop the technology further.

On the other hand, I have been shooting Minolta, and than Sony Alpha 100,700,900 and 55SLT and very seldom found DR an issue, except when sun is included in the image.

Just another observation, DR of digital cameras is wider than any present output media, so some decent tone mapping can be needed to tame the dynamic range.

Best regards
Erik


Negative only, B&W or Color, DR is worst on slides. I wish they would have improved on Fuji's technology further! PHEEWW what a camera this would prove to be with another five years of development! Imagine a 30mpx FF sensor (no need to change the size of either R or S sensors), with further improvement in color, DR and noise! Merry Christmas, Theodoros. www.fotometria.gr
Logged

Pages: « 1 [2] 3 4 ... 8 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad