More megapixels, better photos: Fact or fiction?

Started Feb 7, 2008 | Discussions
John Sheehy Forum Pro • Posts: 21,713
Re: More MP

Simon97 wrote:

I think it is cool when someone sells a photo made with a low pixel
count "obsolete" camera.

This fella make all kinds of photos with various cameras. This one
was made with a basic consumer grade FujiFilm A330.

http://www.pbase.com/purzelbaum/image/32184925

That looks kind of artificial, texture-wise.

I probably wouldn't have made this comment if I came across the image browsing pbase, but since you put it up on a pedestal, I just had to have a go at it.

-- hide signature --

John

Chato
Chato Forum Pro • Posts: 46,792
Re: Well Jay

Jay Turberville wrote:

You can play the same game I'm playing. I post on the Bird Forum:
http://www.birdforum.net/forum.php

They have thousands and thousands of examples of bird photography.
Each time you go there they have random images. I bet with myself,
"was the camera a digcam or an SLR?"

So far I've been correct 100 percent of the time. The digicam
pictures are easy enough to detect...

That's all well and good, but there can be other factors that you are
keying on such as DoF. But regardless, that's not the point you were
making or that I was responding to which had nothing to do with DSLRs
versus compacts. I was responding to your assertion that with
smaller sensors, there is no point in going beyond 5Mp. My
experience is doesn't jibe with that - as I outlined.

Under optimum conditions perhaps - But MY point is that the images I am seeing, images posted by people who believe they are worth posting and viewing - They do indeed look plastic.

Hey, why argue with me? Play my game yourself -

These after all were taken with a $129 camera with 5MP

While this image was taken with a Two MP camera...

Dave

Jay Turberville
http://www.jayandwanda.com

John Sheehy Forum Pro • Posts: 21,713
Re: More megapixels, better photos: Fact or fiction?

newsshooterjim wrote:

The article listed pretty well answers the question. The short of it
is that smaller photosites see less light, this means the signal has
to be amplified more in order to get it up to an acceptable level.
Unfortunately, it's just like a tv amplifier. Boost the signal, you
boost the noise. In low light, my 6mp beats my 10mp K10 every time.

The fact is, the more you divide photons into more numerous, smaller photosites, the less image noise you have to create. The smaller the total number of photons that need to be read out of a sensor, the less noise created as measured in photons. The standard deviation of readout noise in a Canon 5D pixel at ISO 100 with up to 53200 photons is about 30 photons. The standard deviation of readout noise in a Panasonic FZ50 at ISO 100, with a maximum of 4800 photons, is about 3.5 photons. A 5D pixel covers about 16x as much area on a sensor as an FZ50 pixel does. So, 16 FZ50 pixels collect about 4800*16 = 76,800 photons. The read noise for this super pixel is 3.5 photons times the square root of 16, or 4. 3.5*4 = 14 photons of noise. So, the 5D pixel has a noise floor a little over a stop higher than the FZ50 "superpixel", and has slightly more shot noise, with somewhat less photon collection. This is reality; not the myth going around that more pixels makes sensors noisier.

One caveat, some manufacturers are finding ways to work around this,
as you can tell if you've seen the high ISO shots from the D3, and to
a lesser extent, the K20.

The D3 has high-ISO optimizations, like Canon DSLRs have always had except for the original 1D (I don't know much about the D30 and D60). The K20D does not have any high-ISO-specific optimizations, most likely. It most likely follows the K10D paradigm of having low universal read noise, and high ISOs just being a different arithmetic manipulation of the same 22-bit readout. Read noise is exactly 16x as strong at ISO 1600 than it is at ISO 100 on the K10D.

The K20D has slightly less noise than the K10D (both are among the lowest of any digital at ISO 100). It won't touch the D3 or the Canons in the high-ISO department, but it will most likely reign in the low-ISO dynamic range department.

A FF sensor with 13-bit RAW and K20D pixel properties (just 2.25x as many of them) would be an awesome camera for low-ISO uses.

Plenty of light, more megapixels is fine, less light, it's a problem.
For me, the bigger problem is the myopic plastic they're passing off
as lenses.

Buy cameras with good glass, always, regardless of MPs. That's one reason why I like my FZ50, even though it takes over 5 seconds to write a RAW and take the next shot, and the Venus NR is not very nice looking on the JPEGs.

-- hide signature --

John

Amin Sabet
Amin Sabet Veteran Member • Posts: 6,763
Yes and no

In general, for a given level of sensor technology (not comparing old tech to new tech) and a given format size (not comparing small sensors to big sensors), a sensor with more pixels will have more noise in low light and less dynamic range. However, these deficiencies are not as apparent in equal output size prints as they seem based upon examination at the pixel level. Furthermore, a big difference in sensor size is typically associated with a relatively modest difference in dynamic range.

On the other hand, given an adequate lens and sufficient light, the sensor with more pixels generally delivers greater detail in large prints. For small prints, there is no advantage to having lots of pixels, whereas the disadvantages mentioned above may still be evident.

More MP is a tradeoff regardless of format. The same tradeoffs seen with the Fuji F30 and G7 can be seen with the Nikon D3 and Canon 1Ds Mk III.

 Amin Sabet's gear list:Amin Sabet's gear list
Sony a9 Panasonic Lumix DMC-GX85 Nikon D500 Panasonic Lumix G 14mm F2.5 ASPH Panasonic Lumix G Macro 30mm F2.8 +11 more
John Sheehy Forum Pro • Posts: 21,713
Re: I think you are misinformed

sugar wrote:

You only talk about statistics and forget about electronics

1. More pixels will deliver a lower 'net' sensor surface (because of
the circuitery) using the same technology.
2. You assume that the signal/noise ratio per pixel is independent of
pixel size, which is wrong ... google on 'noise floor will learn you
something more about it. If we would continue to make the pixels
smaller, we would arrive at a point were pictures would be a mere
'image' of noise signals.

Now what would your eyes and brain do with information like that?

Think very hard about what I just asked.

-- hide signature --

John

John Sheehy Forum Pro • Posts: 21,713
Re: I believe my eyes

Chato wrote:

DMillier wrote:

Actually, the poster is correct

Sheesh. Everyone seems to think that you have to pixel peep to see
the results of these 10, ,12 MP cameras. I'm referring to looking at
the prints. They are flat, plastic, detail merged into detail...

Isn't the Noise Reduction Race something else?

-- hide signature --

John

John Sheehy Forum Pro • Posts: 21,713
Re: Well Jay

Chato wrote:

While this image was taken with a Two MP camera...

That's a mighty difficult task for an image displayed at 0.3675 MP.

Are you a magician or something?

-- hide signature --

John

Jay Turberville Forum Pro • Posts: 12,917
Re: Well Jay

Under optimum conditions perhaps - But MY point is that the images I
am seeing, images posted by people who believe they are worth posting
and viewing - They do indeed look plastic.

The two cameras have the same sensor size. I simply don't bother with the Coolpix 5000 any more. The similarity of conditions isn't the issue. The 8400 does better in every situation with the possible exception of very low light - where they both do poorly - though the 5000 does slightly less poorly there.

Hey, why argue with me? Play my game yourself -

These after all were taken with a $129 camera with 5MP

While this image was taken with a Two MP camera...

And a swell game that is where we compare apples to tangelos it super small image sizes. I really don't care that much about small web images. I care mostly about prints. And when it comes to prints the 8400 is a little better than the 5000. Oh and while we are at it, my C7070 (7Mp 1/1.8" sensor) does better than my Coolpix 995 (3Mp 1.1.8" sensor). Maybe tomorrow I'll make some proper comparison images for you.

-- hide signature --
John Sheehy Forum Pro • Posts: 21,713
Re: I agree it depends

KenEis wrote:

More pixels do allow better pictures but that assumes all else is
equal. This is never true. Smaller photocites mean more resolution
but also worse noise performance.

Other than the cases where companies improve their efficiencies, yes, but on a per-pixel basis. The noise of an image is not determined by the noise of the pixel alone; the size of the pixel, displayed, is the other factor. Noise is proportional to the standard deviation times the square root of the displayed area of each pixel.

This is something that most analyses of noise in the past have failed to observe.

Small photocites also run into
diffraction limiting at lower and lower f stops.

Small pixels don't make you hit it any faster; that is a myth. Both pixel size and diffraction are parameters that confuse the ideal image. Diffraction makes photons wind up in the wrong place, or confuses their location, and large photosites do the same. Each confuses the photons further.

It seems that the majority of people always have a knack of looking at things the wrong way.

-- hide signature --

John

Amin Sabet
Amin Sabet Veteran Member • Posts: 6,763
Correction

Amin Sabet wrote:

Furthermore, a big
difference in sensor size is typically associated with a relatively
modest difference in dynamic range.

That should read big difference in "pixel size" not "sensor size"

 Amin Sabet's gear list:Amin Sabet's gear list
Sony a9 Panasonic Lumix DMC-GX85 Nikon D500 Panasonic Lumix G 14mm F2.5 ASPH Panasonic Lumix G Macro 30mm F2.8 +11 more
cpw Regular Member • Posts: 313
Re: So many misinformed replies

ejmartin wrote:

...

While the individual parcels have more
variance, since a fixed area of sensor has the same number of photons
falling on it, a fixed area of the image still has the same number of
photons and thus the same sample variance and hence the same noise on
a per area basis. The only thing that more pixels brings is more
resolution, and that's a good thing.

Emil,

Such a general statement as this, i.e. per area noise equality, only strictly holds if there is only shot noise. However, each pixel has read noise on it, it has to taken in to account (I know you talk about read noise below).

To see this, look at an extreme case. Suppose 10,000 e- hits an area patch. One camera has 1 pixel for this area, and a second camera has 10,000 pixels. Both have 20 e- read noise. The SNR for the area in the first camera would be 98. The SNR for the second camera would be about 5, because you have to rss the noise contributions (which in this case is read noise dominated for the second camera) from all of those little pixels.

Now you might say, well I chose a hypothetical situation, which I did, to show this point: that you can't just make this general statement, that per area we get the same noise, irregardless of pixel size, and only use shot noise to prove it.

Chris

misha marinsky3 Senior Member • Posts: 1,330
Re: Correction

For what it's worth, I take IR shots with a 2.1 MP Oly 2000, and a Hoya R72. I can make nice 8.5x11s with it. How many diodes do you need for that or 11x14? How many 11x14 enlargements did anyone make from a 35 negative?

And then there is enlargement software. The more diodes on a given size chip, the faster you reach dimishing returns. Simple physics. Something else: in order to doube the resolution, you have to quadruple the diodes.

http://ubereye.deviantart.com/

Barry Fitzgerald Forum Pro • Posts: 29,888
Re: I agree it depends

John Sheehy wrote:

Small pixels don't make you hit it any faster; that is a myth. Both
pixel size and diffraction are parameters that confuse the ideal
image. Diffraction makes photons wind up in the wrong place, or
confuses their location, and large photosites do the same. Each
confuses the photons further.

It seems that the majority of people always have a knack of looking
at things the wrong way.

It seems the majority of people understand what is going on!

Diffraction becomes more an issue, with more mp..fact. To dispute that, is rather unwise.

As for the noise element..well, we all know your manufacturer sponsored opinions on that one

So far the real world evidence is thin to say the least..

Didnt you crop an SLR image to the same size as an FZ sensor, and declare the FZ sensor to be light years ahead of the SLR? I mean never mind the SLR image is res challenged. Really odd way of working out things, I have to say

-- hide signature --

Clint is on holiday! Soon to return!

ejmartin Veteran Member • Posts: 6,274
Re: So many misinformed replies

cpw wrote:

ejmartin wrote:

...

While the individual parcels have more
variance, since a fixed area of sensor has the same number of photons
falling on it, a fixed area of the image still has the same number of
photons and thus the same sample variance and hence the same noise on
a per area basis. The only thing that more pixels brings is more
resolution, and that's a good thing.

Emil,

Such a general statement as this, i.e. per area noise equality, only
strictly holds if there is only shot noise. However, each pixel has
read noise on it, it has to taken in to account (I know you talk
about read noise below).

Where in my initial post did I talk about shot noise and read noise? I talked about sample variance and sampling error for large and small samples. That those variances and errors have multiple sources doesn't matter so much to a first approximation.

Suppose you are asked to quickly estimate the size of a crowd of people. Do you think that if the size of the crowd is about 10,000 that you will make the same error in your estimate as when the size of the crowd is about 100? The crowd size isn't fluctuating, one just has the analog of read noise here. So read noise varies with pixel size. Shot noise does too, because it quite directly varies with (square root of) sample size.

Why do you think read noise goes down with pixel size in the examples I listed? It's because whatever is counting the electrons has a smaller sample to deal with, and thus makes a smaller counting error.

To see this, look at an extreme case. Suppose 10,000 e- hits an area
patch. One camera has 1 pixel for this area, and a second camera has
10,000 pixels. Both have 20 e- read noise. The SNR for the area in
the first camera would be 98. The SNR for the second camera would be
about 5, because you have to rss the noise contributions (which in
this case is read noise dominated for the second camera) from all of
those little pixels.

If we're going to talk about hypothetical situations, they should reflect reality closely lest wrong conclusions are drawn. That's why in the further discussion I chose measured data from actual production cameras.

-- hide signature --
Chato
Chato Forum Pro • Posts: 46,792
What difficult task?

John Sheehy wrote:

Chato wrote:

While this image was taken with a Two MP camera...

That's a mighty difficult task for an image displayed at 0.3675 MP.

Are you a magician or something?

I have no idea what your point is, but this image, after interpolating up to plus 100MP prints out just fine at 16x20.

Of course there's not much detail in the image - So what? Olympus 510.

Dave

John Sheehy Forum Pro • Posts: 21,713
Re: I agree it depends

Barry Fitzgerald wrote:

John Sheehy wrote:

Small pixels don't make you hit it any faster; that is a myth. Both
pixel size and diffraction are parameters that confuse the ideal
image. Diffraction makes photons wind up in the wrong place, or
confuses their location, and large photosites do the same. Each
confuses the photons further.

It seems that the majority of people always have a knack of looking
at things the wrong way.

It seems the majority of people understand what is going on!

Diffraction becomes more an issue, with more mp..fact. To dispute
that, is rather unwise.

Diffraction becomes a more significant limiter (only in terms of relative contribution to confusion) of overall MTF as the pixel density reaches a certain point. Increasing pixel density NEVER causes diffraction to be more of a problem than with a lower pixel density.

Like 99% of the world, you are looking at things backwards. Diffraction issues are not a matter of what f-stop a pixel-pitch becomes "diffraction limited"; the real issue is how much diffraction lowers the overall MTF, and the higher the pixel density, the less that pixel density contributes to overall loss of MTF.

This is so simple once you see it this way; the way it really is - a series of obstacles that each throw the photons a little bit more off from a direct course.

As for the noise element..well, we all know your manufacturer
sponsored opinions on that one

I have demonstrated time and again that one factor in noise is the displayed pixel size. No one has ever presented any real evidence that higher pixel densities are detrimental to IQ. It's a myth, like the flat earth.

So far the real world evidence is thin to say the least.

Good evidence that you choose to call thin because you don't want to believe it; however, there is no evidence to support what you believe, whatsoever, except circumstantial evidence involving rising expectations, overly fond and unrealistic memories of cameras past, horrible display options, and artifacts of the NR race.

Didnt you crop an SLR image to the same size as an FZ sensor, and
declare the FZ sensor to be light years ahead of the SLR?

Aren't you one of the people who didn't understand the significance of that?

I mean
never mind the SLR image is res challenged. Really odd way of working
out things, I have to say

Not really, when you consider that more detail of the subject was visible, and noise was about the same in both.

-- hide signature --

John

John Sheehy Forum Pro • Posts: 21,713
Re: What difficult task?

Chato wrote:

I have no idea what your point is, but this image, after
interpolating up to plus 100MP prints out just fine at 16x20.

Of course there's not much detail in the image - So what? Olympus 510.

My point is that I don't see your point; a 2 MP image displayed at 0.3675 MP - what is that supposed to tell us?

-- hide signature --

John

Chato
Chato Forum Pro • Posts: 46,792
Whatever you want...

John Sheehy wrote:

My point is that I don't see your point; a 2 MP image displayed at
0.3675 MP - what is that supposed to tell us?

Nice picture?

Dave

Eric Fossum
Eric Fossum Senior Member • Posts: 1,380
Re: More megapixels, better photos: Fact or fiction?

John Sheehy wrote:

The fact is, the more you divide photons into more numerous, smaller
photosites, the less image noise you have to create. The smaller the
total number of photons that need to be read out of a sensor, the
less noise created as measured in photons. The standard deviation of
readout noise in a Canon 5D pixel at ISO 100 with up to 53200 photons
is about 30 photons. The standard deviation of readout noise in a
Panasonic FZ50 at ISO 100, with a maximum of 4800 photons, is about
3.5 photons. A 5D pixel covers about 16x as much area on a sensor as
an FZ50 pixel does. So, 16 FZ50 pixels collect about 4800*16 =
76,800 photons. The read noise for this super pixel is 3.5 photons
times the square root of 16, or 4. 3.5*4 = 14 photons of noise. So,
the 5D pixel has a noise floor a little over a stop higher than the
FZ50 "superpixel", and has slightly more shot noise, with somewhat
less photon collection. This is reality; not the myth going around
that more pixels makes sensors noisier.

No myth John, if I understand what you are saying correctly. Noise has many components, including photon shot noise that goes like the sqrt of the number of carriers, and readout noise, which is usually fixed as a function of signal. We usually care about signal to noise ratio, SNR. So for shot noise, S divided by sqrt (S) = sqrt(S). So while the noise is less, when people say "noiseier" they are almost always talking about SNR, which degrades with smaller pixel sizes for the most part.

Did I understand you correctly?
-Eric

 Eric Fossum's gear list:Eric Fossum's gear list
Nikon Coolpix P900 Sony RX100 II
DRG Veteran Member • Posts: 5,217
Re: More megapixels, better photos: Fact or fiction?

Eric Fossum wrote:

No myth John, if I understand what you are saying correctly. Noise
has many components, including photon shot noise that goes like the
sqrt of the number of carriers, and readout noise, which is usually
fixed as a function of signal. We usually care about signal to noise
ratio, SNR. So for shot noise, S divided by sqrt (S) = sqrt(S). So
while the noise is less, when people say "noiseier" they are almost
always talking about SNR, which degrades with smaller pixel sizes for
the most part.

Did I understand you correctly?

Based on what John's written in prior posts, I believe he's concerned with the overall noise in the image (with the implication that any two images should be compared at the same output size) and not with the SNR on a per pixel basis. The ultimate basis for his argument is that a sensor that could enumerate the exact location of each and every photon hit would be capturing the maximum information possible in an image and that the output of any conventional finite pixel count sensor could be derived from that data and thus cannot exceed the quality of the data. So, more pixels equals better from an overall image point of view when comparing images at the same output size, on a hypothetical monitor (or other output medium) capable of displaying all the included information rather than resorting to downsampling. Oh, and with the additional assumption that the total photosensitive area is the same in all cases (that the electronics are somehow scaling to take up the same fraction of the pixel area).

I probably shouldn't be speaking for John but, hey, I have free wi-fi access in Starbucks and feel the need to take advantage of it

David

Keyboard shortcuts:
FForum MMy threads