Pixel Density is GENIUS!

Started Jul 13, 2008 | Discussions
RRJackson
OP RRJackson Senior Member • Posts: 2,555
Re: No really, it isn't

simpy wrote:

Is it that much noisier if you downsample both to the same
resolution? It may be, of course, but I haven't seen any evidence to
think it is. If it is, it goes on the pile of anecdotal evidence
indeed, but please forgive me in pointing out that you have also
conveniently ignored the image samples by John Sheehy that would go
on the pile of counterevidence.

I haven't seen any test charts from Mr. Sheehy. I've seen an underexposed low-contrast snapshot that he claims came from a sensor like the one in the Fuji, but refused to expound on the source of the sensor. I've also seen some underexposed crops from two captures that he claims as evidence that smaller photosites are superior to larger photosites in low light. Neither of these presentations has been particularly well documented or controlled.

Meanwhile this site is a massive repository of camera test data that anecdotally points to pixel density as a factor.

I don't have a problem with pixel density not being a factor. It's not like I build a low-pixel-density compact, or even own one. Although I gave my girlfriend a little Panasonic a couple of years ago. My only problem with it conceptually is that I haven't seen any practical evidence of it. Or at least nothing that struck me as being particularly thorough or convincing.

ejmartin Veteran Member • Posts: 6,274
Re: No really, it isn't

RRJackson wrote:

simpy wrote:

Is it that much noisier if you downsample both to the same
resolution? It may be, of course, but I haven't seen any evidence to
think it is. If it is, it goes on the pile of anecdotal evidence
indeed, but please forgive me in pointing out that you have also
conveniently ignored the image samples by John Sheehy that would go
on the pile of counterevidence.

I haven't seen any test charts from Mr. Sheehy. I've seen an
underexposed low-contrast snapshot that he claims came from a sensor
like the one in the Fuji, but refused to expound on the source of the
sensor. I've also seen some underexposed crops from two captures that
he claims as evidence that smaller photosites are superior to larger
photosites in low light. Neither of these presentations has been
particularly well documented or controlled.

Meanwhile this site is a massive repository of camera test data that
anecdotally points to pixel density as a factor.

I have a little exercise for you:

1. Open a blank canvas in Photoshop. Fill with a uniform middle gray tone. (Image> Adjustments> Levels and set the output level to 128)

2. Add some gaussian noise (Filters> Noise> Add Noise... and choose some amount, say 10%). Look at the histogram and note how its width has broadened from a sharp spike at 128 into a bell curve by the addition of the noise. The width of this bell curve measures the amount of noise.

3. Open the Gaussian blur filter (Filters> Blur> Gaussian Blur...). Watch what happens to the histogram as you very the radius of the blur. You will see the width of the noise histogram decrease in proportion to the radius of the blur. Since the width is a quantitative measure of the amount of noise, the noise decreases as the blur is increased.

Lest you get confused about what you should infer from this exercise, let me spell it out to you. The blur in step 3 decreases resolution by averaging over pixels. Noise decreases in proportion to the amount of resolution decrease. If one downsamples an image properly, one decreases the resolution, and noise decreases in proportion to the linear change in image size.

There is little difference in this regard between shrinking the size of the image keeping the pixel size fixed -- which reduces the resolution by the percentage of shrinkage -- versus keeping the image size fixed and applying a blur, which again reduces the resolution by the ratio of the blur diameter to the pixel spacing. Either way one is averaging over pixels of the original image, and that reduces noise.

The conclusion is that noise is resolution dependent. If you compare two cameras having the same sensor size and technologies but different resolutions (pixel densities), the one with the lower resolution (lower pixel density) will have less noise at the pixel level, as the above exercise has demonstrated.

This site always tests cameras at 100% pixel resolution, and so noise comparisons between cameras of different pixel density, even with the same sensor size, are skewed in favor of the one with lower pixel density, because of the result of the above exercise. Properly resample the higher pixel density to the pixel dimensions of the lower pixel density one, and the comparison becomes fair -- and moreover is a truer indication of what you will see when viewing prints from the cameras at the same distance.

For a quick rule of thumb for comparing cameras of different pixel density fairly, divide the noise measurements of each (as they are done on this site) by the square root of the MP count, which is the linear resolution relative to image size. Note that the noise figure of the higher MP camera will be divided by a larger number, and so come into line with the lower MP camera as a result.

When one does this exercise, it becomes apparent that the main factor in image noise is sensor size. The result is largely independent of MP count for a fixed sensor size. Pixel density, which is sensor area divided by MP count, is poorly correlated to noise because both MP count and sensor size will vary from camera to camera, but only one of those factors is tied to noise level.

-- hide signature --
RRJackson
OP RRJackson Senior Member • Posts: 2,555
Re: No really, it isn't

ejmartin wrote:

I have a little exercise for you:

1. Open a blank canvas in Photoshop.

Yes...sure. That's the song that's been sung since Phil announced the spec was going to be part of the site.

When one does this exercise, it becomes apparent that the main factor
in image noise is sensor size. The result is largely independent of
MP count for a fixed sensor size. Pixel density, which is sensor
area divided by MP count, is poorly correlated to noise because both
MP count and sensor size will vary from camera to camera, but only
one of those factors is tied to noise level.

This assumed a resampled output size. If you were to always print your output at 300 dpi and just allow the print to be whatever size it works out to be then the noise becomes a constant and not a variable.

simpy Veteran Member • Posts: 3,090
Re: No really, it isn't

RRJackson wrote:

I haven't seen any test charts from Mr. Sheehy. I've seen an
underexposed low-contrast snapshot that he claims came from a sensor
like the one in the Fuji, but refused to expound on the source of the
sensor. I've also seen some underexposed crops from two captures that
he claims as evidence that smaller photosites are superior to larger
photosites in low light. Neither of these presentations has been
particularly well documented or controlled.

I presume that you mean this post http://forums.dpreview.com/forums/read.asp?forum=1018&message=28607494
I think it is fairly well controlled, as far as these things go.

Meanwhile this site is a massive repository of camera test data that
anecdotally points to pixel density as a factor.

That is possibly the biggest problem of all. This site only reports on noise at the pixel level. It is obvious that this will give you higher noise for smaller pixels, but that doesn't mean it is relevant. What a user should really care about is the noise at a given output size, and I and others claim that this is not strongly dependent on sensor size at all.

My only problem with it conceptually is that I haven't seen any
practical evidence of it. Or at least nothing that struck me as being
particularly thorough or convincing.

Let me turn it around. Have you seen any thorough or convincing evidence that increasing pixel density does increase noise for a given output size?

Simon

 simpy's gear list:simpy's gear list
Olympus E-M1 Olympus M.Zuiko Digital ED 9-18mm F4.0-5.6 Olympus M.Zuiko Digital 45mm F1.8 Olympus M.Zuiko Digital 17mm F1.8 Olympus Zuiko Digital ED 50mm 1:2.0 Macro +7 more
RRJackson
OP RRJackson Senior Member • Posts: 2,555
Re: No really, it isn't

simpy wrote:

Let me turn it around. Have you seen any thorough or convincing
evidence that increasing pixel density does increase noise for a
given output size?

Simon

Sure. Native output size.

simpy Veteran Member • Posts: 3,090
Re: No really, it isn't

RRJackson wrote:

This assumed a resampled output size. If you were to always print
your output at 300 dpi and just allow the print to be whatever size
it works out to be then the noise becomes a constant and not a
variable.

Please explain why you'd want to do that. The only reason I can think of is that for some reason you'd like to print as big as possible and you're allergic to upsampling. Then you are limited by the camera resolution, so we're back in a more MP=higher pixel density=better scenario.

That's not what you were arguing, right? Neither am I. The reasonable thing is to compare images at the same output size.

Simon

 simpy's gear list:simpy's gear list
Olympus E-M1 Olympus M.Zuiko Digital ED 9-18mm F4.0-5.6 Olympus M.Zuiko Digital 45mm F1.8 Olympus M.Zuiko Digital 17mm F1.8 Olympus Zuiko Digital ED 50mm 1:2.0 Macro +7 more
simpy Veteran Member • Posts: 3,090
Re: No really, it isn't

RRJackson wrote:

Sure. Native output size.

What is so magical about 'native' output size? Please explain, because I really have a hard time understanding why this would be relevant.

Simon

 simpy's gear list:simpy's gear list
Olympus E-M1 Olympus M.Zuiko Digital ED 9-18mm F4.0-5.6 Olympus M.Zuiko Digital 45mm F1.8 Olympus M.Zuiko Digital 17mm F1.8 Olympus Zuiko Digital ED 50mm 1:2.0 Macro +7 more
RRJackson
OP RRJackson Senior Member • Posts: 2,555
Re: No really, it isn't

simpy wrote:

What is so magical about 'native' output size? Please explain,
because I really have a hard time understanding why this would be
relevant.

Do you scale video down to watch it so the chroma noise will be less objectionable?

The usefulness of a camera, IMO, is largely dependent on it's largest printable output, or native resolution. All of which are lacking until you get into very expensive digital backs. A 1DsIII at 300 dpi yields a print 12 1/2" x 18 3/4" (more or less). You're not going to be giving Andreas Gursky a run for his money at those kinds of resolutions. So right now there's no "spare" resolution that can be tossed away in the interest of lowering the noise floor. If it looks bad at its native resolution then it's not a viable product for professional use. Of course, these tiny sensors aren't being used by professionals for the most part. But the validity of pixel density still seems compelling.

simpy Veteran Member • Posts: 3,090
skewed comparison

RRJackson wrote:

Do you scale video down to watch it so the chroma noise will be less
objectionable?

[snip
If it looks
bad at its native resolution then it's not a viable product for
professional use. Of course, these tiny sensors aren't being used by
professionals for the most part. But the validity of pixel density
still seems compelling.

This comparison really doesn't fly when we are discussing sensors of the same size. The sensor with lower pixel density has less noise at the pixel level, but has low resolution limiting it in output size. The sensor with higher pixel density has more resolution, which means it could be used for sharp prints of larger size. However, the enlargement potential could be limited by noise, but when printed to the same size, the quality would not be worse than the low pixel density sensor. In this case, higher pixel density gives you more flexibility.

Please indicate if you agree with the reasoning in the paragraph above, and if not, exactly what it is you don't agree with.

In addition, you seem to assume that 'professional use' equates to printing as large as possible. Although this could apply to specific niches (you mentioned Gursky), it certainly isn't the case in general.

Simon

 simpy's gear list:simpy's gear list
Olympus E-M1 Olympus M.Zuiko Digital ED 9-18mm F4.0-5.6 Olympus M.Zuiko Digital 45mm F1.8 Olympus M.Zuiko Digital 17mm F1.8 Olympus Zuiko Digital ED 50mm 1:2.0 Macro +7 more
John Sheehy Forum Pro • Posts: 21,841
Re: No really, it isn't

RRJackson wrote:

It's interesting to see what you were able to do with it. I still
don't think you're going to convince many pros to dump their 1DIII
and start shooting with a G9.

Huh? the context here would be something like one of the Fuji P&S against the G9, not a DSLR vs the G9.

-- hide signature --

John

John Sheehy Forum Pro • Posts: 21,841
Re: No really, it isn't

RRJackson wrote:

Well, that is useful. Because given the choice of getting 12
megapixels from a 1/1.7" sensor or a 135 sensor I'd much rather have
the lower pixel density of the 135 sensor.

Yes, and you'd be in error if you thought the camera had less noise because of the lower pixel density. The fact is, the sensor is a lot larger.

But sticking with like-sized sensors, I notice that the 1DsIII is a
lot noisier at high ISO than the D3.

Since you obviously can not tell the difference between low noise and noise reduction, that's not surprising. The two cameras have similar image-level noise in the highlights, and the 1Ds3 has less image-level read noisie in the shadows, in the RAW data.

Another one of those unfortunate
coincidences that keeps piling up on the mountain of anecdotal
evidence?

None of which is from RAW data, always from unequally noise-reduced samples. The real RAW story contradicts your claim.

-- hide signature --

John

RRJackson
OP RRJackson Senior Member • Posts: 2,555
Re: skewed comparison

simpy wrote:

This comparison really doesn't fly when we are discussing sensors of
the same size. The sensor with lower pixel density has less noise at
the pixel level, but has low resolution limiting it in output size.
The sensor with higher pixel density has more resolution, which means
it could be used for sharp prints of larger size. However, the
enlargement potential could be limited by noise, but when printed to
the same size, the quality would not be worse than the low pixel
density sensor. In this case, higher pixel density gives you more
flexibility.

Please indicate if you agree with the reasoning in the paragraph
above, and if not, exactly what it is you don't agree with.

I don't know how well this holds up theoretically, but practically I used to print down a size to minimize grain. I basically had to halve the print size to gain a stop of "NR" so I'd go from an 8x10 to a 5x7 if I wanted the image, but wasn't happy with the prominence of grain. It didn't seem to make a noticeable (or rather practical) difference until I'd halved the print size (roughly, since a 5x7 is actually a little less than half). If this holds true with digital noise (and I haven't printed much of my own digital output...I have a little dye-sub for quick prints, but I send out for enlargements) then the jump from (for example) 8 to 12 megapixels would be a losing game. It seems like you have to be talking about a sensor with twice the density. At which point the level of light where the sensor's output became unusable even with scaling would be lower, as well. I don't see any advantages to it. That's not to say that there aren't any, but I can't see them from a practical sense.

The game changed a lot over the last five years or so and it still seems like we're seeing a stop or so of improvement every couple of years. Whether that's through sensor efficiency or through NR "cheating" is another matter. But I'd much rather use a sensor that's suited well to a wide variety of operating conditions than one that can be scaled and blurred and noise-reduced enough to be kind of acceptable if you make it small enough.

In addition, you seem to assume that 'professional use' equates to
printing as large as possible. Although this could apply to specific
niches (you mentioned Gursky), it certainly isn't the case in general.

I printed larger than 8x10 three times last year. Had to think about it. I used to do it a lot more often, but I'm not that active right now. So I'll certainly admit that printing large isn't an everyday occurrence for me. But then we need to decide whether we're looking for technology that's optimal or passable.

RRJackson
OP RRJackson Senior Member • Posts: 2,555
Re: No really, it isn't

John Sheehy wrote:

None of which is from RAW data, always from unequally noise-reduced
samples. The real RAW story contradicts your claim.

You make a lot of claims, but it seems like when you try to substantiate them you resort to things like comparing a tiny slice of the optics of an APS-sensor camera to an optimal slice of a point and shoot, so I have a hard time taking your claims at face value.

Roland Karlsson Forum Pro • Posts: 28,224
Re: No really, it isn't

RRJackson wrote:

You make a lot of claims, but it seems like when you try to
substantiate them you resort to things like comparing a tiny slice of
the optics of an APS-sensor camera to an optimal slice of a point and
shoot, so I have a hard time taking your claims at face value.

This is the main problem when doing the comparisons.

The sensor size varies much more than the pixel count.

The pixel count is 8-16 MP no matter how big the sensor is. Thats a factor two - and the technology used affects the quality more.

So - if you want to have widely different pixel density to get conclusive measurements you have to crop the larger sensors in order to do the comparison.

-- hide signature --

Roland

 Roland Karlsson's gear list:Roland Karlsson's gear list
Sigma DP3 Merrill Sigma dp2 Quattro Sony RX100 III Pentax K-3 Pentax K-1 +14 more
ejmartin Veteran Member • Posts: 6,274
Re: No really, it isn't

RRJackson wrote:

simpy wrote:

What is so magical about 'native' output size? Please explain,
because I really have a hard time understanding why this would be
relevant.

Do you scale video down to watch it so the chroma noise will be less
objectionable?

If video camera manufacturers all made independent independent decisions as to what resolution to offer, and the viewer had a choice in resolution of display devices (beyond the standard/HD dichotomy), then indeed video display devices would have to resample their inputs for the viewer, and one would be faced with precisely the same issues. Print size one already has an analog of -- how big a screen do you want? But also one would have a choice of resolutions in both screens and cameras, and the market would be in as much of a muddle as the camera market.

But the video industry decided for both their convenience and that of the consumer to enforce a standard, and so there are a limited number of resolutions relative to frame size, and so the consumer only has a choice of frame size for viewing.

Now, Phil wants a camera industry standard similar to that of the video industry -- all cameras being 6MP, independent of sensor size. Pretty silly, given that image quality has more to do with sensor size than MP count.

The usefulness of a camera, IMO, is largely dependent on it's largest
printable output, or native resolution.

The native resolution has nothing to do with print size, they are only related when one specifies the printer resolution, and even then only if one ignores the possibility to resample.

What is the logic of the argument? That camera A with higher MP but the same image noise as camera B, is less capable because it gives the same output as camera B upon downsampling, but in addition allows the same image noise and more resolution without resampling (or upon upsampling camera B)?

-- hide signature --
Andrew dB Contributing Member • Posts: 970
Re: No really, it isn't

John Sheehy wrote:

RRJackson wrote:

It's interesting to see what you were able to do with it. I still
don't think you're going to convince many pros to dump their 1DIII
and start shooting with a G9.

Huh? the context here would be something like one of the Fuji P&S
against the G9, not a DSLR vs the G9.

Is it Barry Fitzgerald in disguise?

RRJackson
OP RRJackson Senior Member • Posts: 2,555
Re: No really, it isn't

ejmartin wrote:

Now, Phil wants a camera industry standard similar to that of the
video industry -- all cameras being 6MP, independent of sensor size.

Funny, I seem to have missed that press release altogether.

simpy Veteran Member • Posts: 3,090
Re: skewed comparison

RRJackson wrote:

I don't know how well this holds up theoretically, but practically I
used to print down a size to minimize grain. I basically had to halve
the print size to gain a stop of "NR" so I'd go from an 8x10 to a 5x7
if I wanted the image, but wasn't happy with the prominence of grain.
It didn't seem to make a noticeable (or rather practical) difference
until I'd halved the print size (roughly, since a 5x7 is actually a
little less than half). If this holds true with digital noise (and I
haven't printed much of my own digital output...I have a little
dye-sub for quick prints, but I send out for enlargements) then the
jump from (for example) 8 to 12 megapixels would be a losing game. It
seems like you have to be talking about a sensor with twice the
density. At which point the level of light where the sensor's output
became unusable even with scaling would be lower, as well. I don't
see any advantages to it. That's not to say that there aren't any,
but I can't see them from a practical sense.

Printing at smaller sizes to suppress grain (noise) is indeed exactly equivalent to downsampling.

As to whether it is better or not to start with extra pixels, if you do the math, you'll see that it actually makes no difference (assuming constant read noise per unit area).

But I'd much rather use a sensor that's
suited well to a wide variety of operating conditions than one that
can be scaled and blurred and noise-reduced enough to be kind of
acceptable if you make it small enough.

No, the more flexible sensor is the one with more pixels. It can be made exactly as 'acceptable' as the low pixel count sensor at smaller print sizes, but it has the additional benefit of using the extra resolution when needed.

I printed larger than 8x10 three times last year. Had to think about
it. I used to do it a lot more often, but I'm not that active right
now. So I'll certainly admit that printing large isn't an everyday
occurrence for me. But then we need to decide whether we're looking
for technology that's optimal or passable.

Optimal, optically speaking, would be a sensor that is as large as possible, because it gives you most flexibility. In practice, you'll always get what's passable for your particular needs and budget (money, size, weight, etc.).

I think I need to clarify one thing, though. The pixel density can be calculated from the size and MP count. What you seem to be arguing is that at a given MP count, decreasing the pixel density leads to better quality through the larger sensor size. No argument from me, but the fact that it is clearer to talk about the sensor size directly.

The rest of this thread is about what happens if you keep the sensor size constant. In that case, is increasing the MP count a good thing or a bad thing? My point is it definitely doesn't need to be a bad thing.

Simon

 simpy's gear list:simpy's gear list
Olympus E-M1 Olympus M.Zuiko Digital ED 9-18mm F4.0-5.6 Olympus M.Zuiko Digital 45mm F1.8 Olympus M.Zuiko Digital 17mm F1.8 Olympus Zuiko Digital ED 50mm 1:2.0 Macro +7 more
RRJackson
OP RRJackson Senior Member • Posts: 2,555
Re: skewed comparison

simpy wrote:

The rest of this thread is about what happens if you keep the sensor
size constant. In that case, is increasing the MP count a good thing
or a bad thing? My point is it definitely doesn't need to be a bad
thing.

And I'd be the first to applaud a practical implementation of that I'd never turn my nose up at extra resolution. I'm just not convinced that the more "pixel-dense" sensor won't hit an unusably high noise floor at light levels where the less "pixel-dense" sensor was still functioning acceptably.

simpy Veteran Member • Posts: 3,090
Re: skewed comparison

RRJackson wrote:

And I'd be the first to applaud a practical implementation of that
I'd never turn my nose up at extra resolution. I'm just not convinced
that the more "pixel-dense" sensor won't hit an unusably high noise
floor at light levels where the less "pixel-dense" sensor was still
functioning acceptably.

To be clear, I'm absolutely fine with your 'not being convinced'. I took objection to your opening statement in this thread: "We've suffered at the hands of the "Megapixel Race" for years now." I hope the discussion has made it clear that there is a lot more to it than pixel density = bad.

Simon

 simpy's gear list:simpy's gear list
Olympus E-M1 Olympus M.Zuiko Digital ED 9-18mm F4.0-5.6 Olympus M.Zuiko Digital 45mm F1.8 Olympus M.Zuiko Digital 17mm F1.8 Olympus Zuiko Digital ED 50mm 1:2.0 Macro +7 more
Keyboard shortcuts:
FForum MMy threads