How does "total light" change SNR? II

[Great Bustard]

The last thread closed out with quite a few ignorant comments that I'd like to address:

Light is finite and fixed and is proportional to sensor area. Resampling has got nothing to do with the amount, it only affects resolution. You can resample a FF sensor as many times as you want but it will not change light. Subdividing the area does not change the total amount of light and yet equivalence is treating it as if it does. This is why it is completely bullcrap.

Actually, the bullcrap is saying that Equivalence says that sampling the signal with more pixels changes the total amount of light. In fact, Equivalence says the exact opposite -- that the total amount of light that makes up the photo is independent of the pixel count.

Futhermore, statements from the same poster in a previous post, such as:

This kind of argument shows lack of understanding of sampling theory. Same sensel density results in the same sampling frequency.

show that he got the two sentences exactly backwards. Specifically, the same pixel density on sensors with different sizes results in different sampling frequencies since the natural and implicit conditions for the comparison are the same perspective and framing.

The above poster's conclusion does not end there. He closes his post with:

The only conclusion and the only correct conclusion is that sensor size has absolutely got nothing to do with photographic exposure and therefore total light is meaningless when not associated with total sensor area. Photography has always been and always will be LIGHT SPREAD OVER AN AREA. Ultimately, noise is the same regardless of format.

which demonstrates a *gross* misunderstanding of both Equivalence and photography. First of all, no one says, or implies, that sensor size has anything to do with exposure except inasmuch as the same DOF and shutter speed will necessarily result in a lower exposure on the larger sensor.

Furthermore, photography has always been about the total amount of light that makes up the photo, not the amount of light per area.

Lastly, I would be hard pressed to find a self-harming post where anyone has said anything as stupid as "Ultimately, noise is the same regardless of format." Even in the case of Equivalent photos, which results in the same total amount of light falling on the sensor, the noise also depends heavily on the sensor efficiency.

 

[D Cox]

The last thread closed out with quite a few ignorant comments that I'd like to address:

Light is finite and fixed and is proportional to sensor area. Resampling has got nothing to do with the amount, it only affects resolution. You can resample a FF sensor as many times as you want but it will not change light. Subdividing the area does not change the total amount of light and yet equivalence is treating it as if it does. This is why it is completely bullcrap.

Actually, the bullcrap is saying that Equivalence says that sampling the signal with more pixels changes the total amount of light. In fact, Equivalence says the exact opposite -- that the total amount of light that makes up the photo is independent of the pixel count.

True.

And the amplitude of noise in the raw image is independent of the "total light".

Are you the person who invented this concept of "total light" ?

 

[DSPographer]

After reading through some of the posts in the previous thread it became obvious that a few posters were fixated on the results of how they measured noise. So, noticing how measuring the standard deviation of pixels for a uniform patch changes according to pixel size on the sensor, or how it changes by averaging of pixels leads to misunderstandings about what noise is and how it is affected by pixel size.

When we look at an image we don't just average the level of all or a group of pixels, we see details as a function of how large they are with respect to the image as viewed. Small grained image noise and large grained noise are perceived separately, just as the image details are. So what really matters, when looking at how noisy an image appears, is the noise as a function of its size as a fraction of the total image, compared to the image details by size as a fraction of the total image.

Not appreciating this fact seems to be at the root of all the misunderstandings in the previous thread. This is because if we convert measurements at the pixel or lp/mm level to measurements normalized to image fraction, then if we do the math correctly, how noise scales by total light becomes obvious.

 

[Great Bustard]

The last thread closed out with quite a few ignorant comments that I'd like to address:

Light is finite and fixed and is proportional to sensor area. Resampling has got nothing to do with the amount, it only affects resolution. You can resample a FF sensor as many times as you want but it will not change light. Subdividing the area does not change the total amount of light and yet equivalence is treating it as if it does. This is why it is completely bullcrap.

Actually, the bullcrap is saying that Equivalence says that sampling the signal with more pixels changes the total amount of light. In fact, Equivalence says the exact opposite -- that the total amount of light that makes up the photo is independent of the pixel count.

True.

And the amplitude of noise in the raw image is independent of the "total light".

Is that so? If we take two photos of the same scene with the same camera and lens, one at f/2.8 1/100 and the other at f/5.6 1/100, and display them at the same size and same brightness, which is more noisy and why?

Are you the person who invented this concept of "total light" ?

No. I'm the person who coined the term "total light" to describe the total amount of light falling on the sensor while the shutter is open. Alternatively, one can think in terms of "equivalent exposure" where the "equivalent exposure" is the amount of light that falls on the sensor for a given proportion of the photo. However, I felt the term "equivalent exposure" lent itself to misinterpretations, so I avoided using it.

 

[D Cox]

And the amplitude of noise in the raw image is independent of the "total light".

Is that so? If we take two photos of the same scene with the same camera and lens, one at f/2.8 1/100 and the other at f/5.6 1/100, and display them at the same size and same brightness, which is more noisy and why?

The greater exposure gives more photons on each pixel, reducing the error in measurement. Therefore there is less noise across the image.

This also applies to the highlight areas of the image as compared to the shadows. But changing the proportion of the image that is occupied by highlights does not affect the noise levels in either shadows or highlights.

Are you the person who invented this concept of "total light" ?

No. I'm the person who coined the term "total light" to describe the total amount of light falling on the sensor while the shutter is open. Alternatively, one can think in terms of "equivalent exposure" where the "equivalent exposure" is the amount of light that falls on the sensor for a given proportion of the photo. However, I felt the term "equivalent exposure" lent itself to misinterpretations, so I avoided using it.

Unfortunately "total light" is just as misleading.

 

[bobn2]

And the amplitude of noise in the raw image is independent of the "total light".

Is that so? If we take two photos of the same scene with the same camera and lens, one at f/2.8 1/100 and the other at f/5.6 1/100, and display them at the same size and same brightness, which is more noisy and why?

The greater exposure gives more photons on each pixel, reducing the error in measurement.

I asked this before, never got an answer. How, exactly, does having more photons 'reduce the error in measurement'? It would be a good start to explain first what you think it is being measured.

 

[gollywop]

I'm beginning to realize that this dtmateojr guy (Demosthenes Mateo Jr.) is simply in seventh heaven playing in the sandboxes he creates here on dpr. After observing his behavior in a couple of recent threads, I became curious to figure out what might be motivating it, and I believe that it that he is just plainly overjoyed at having finally found a way to get people to talk to him. He needs attention at any cost.

It is of interest to note, for example, that his twitter page, which began in May 2011, has 69 tweets and absolutely zero replies over two years. His last tweet was July 6, 2013, shortly after his joining dpr on June 13, 2013. In this new venue he has clearly found a way to attract attention, and as long as he continues to get it, I am convinced he will just go on and on with his current purposely inane and obtuse behavior.

Bear this in mind when you consider responding to his posts. He isn't going to change. He has no reason to. He has what he wants/needs, and you will just be feeding its continuation.

You will also note that the comments made to his famously fatuous post, to which he frequently refers, are best summed up with reilly diefenbach's lead-off comment: Possibly the silliest post having to do with photography anyone has ever made.

He is apparently also (are you ready for this?) a RedHat Certified Engineer, an accomplishment for which he is inordinately proud. You can glean another dimension of his personality from his response to the question as to what being a RHCE means to him. You will note that all the others on the page give responses that actually answer that question. dtmateojr's response, however, is nothing but a self-serving brag session that indicates that the only meaning for him of being a RHCE is in its providing an exam or two on which he could get perfect scores.

Play in his sandbox if you will, but recognize what's buried there.

--
gollywop

 

[bobn2]

I'm beginning to realize that this dtmateojr guy (Demosthenes Mateo Jr.) is simply in seventh heaven playing in the sandboxes he creates here on dpr. After observing his behavior in a couple of recent threads, I became curious to figure out what might be motivating it, and I believe that it that he is just plainly overjoyed at having finally found a way to get people to talk to him. He needs attention at any cost.

It is of interest to note, for example, that his twitter page, which began in May 2011, has 69 tweets and absolutely zero replies over two years. His last tweet was July 6, 2013, shortly after his joining dpr on June 13, 2013. In this new venue he has clearly found a way to attract attention, and as long as he continues to get it, I am convinced he will just go on and on with his current purposely inane and obtuse behavior.

Bear this in mind when you consider responding to his posts. He isn't going to change. He has no reason to. He has what he wants/needs, and you will just be feeding its continuation.

You will also note that the comments made to his famously fatuous post, to which he frequently refers, are best summed up with reilly diefenbach's lead-off comment: Possibly the silliest post having to do with photography anyone has ever made.

He is apparently also (are you ready for this?) a RedHat Certified Engineer, an accomplishment for which he is inordinately proud. You can glean another dimension of his personality from his response to the question as to what being a RHCE means to him. You will note that all the others on the page give responses that actually answer that question. dtmateojr's response, however, is nothing but a self-serving brag session that indicates that the only meaning for him of being a RHCE is in its providing an exam or two on which he could get perfect scores.

Play in his sandbox if you will, but recognize what's buried there.

 

[bobn2]

After reading through some of the posts in the previous thread it became obvious that a few posters were fixated on the results of how they measured noise. So, noticing how measuring the standard deviation of pixels for a uniform patch changes according to pixel size on the sensor, or how it changes by averaging of pixels leads to misunderstandings about what noise is and how it is affected by pixel size.

When we look at an image we don't just average the level of all or a group of pixels, we see details as a function of how large they are with respect to the image as viewed. Small grained image noise and large grained noise are perceived separately, just as the image details are. So what really matters, when looking at how noisy an image appears, is the noise as a function of its size as a fraction of the total image, compared to the image details by size as a fraction of the total image.

Not appreciating this fact seems to be at the root of all the misunderstandings in the previous thread. This is because if we convert measurements at the pixel or lp/mm level to measurements normalized to image fraction, then if we do the math correctly, how noise scales by total light becomes obvious.

The problem really is tying down the perceptual importance of easily measured metrics. Measurements on the scale of a pixel have become second nature to sensor engineers, because they do provide some easily produced figures of merit that inform the development process. They aren't always straightforwardly applicable to perceptual quality, and there hasn't been a whole load of thought through perceptual work done (it's actually not easy to do). But, as a first approximation, I tend to go with the notion that the more information that is captured about the image projected on the sensor, the greater the likelihood of producing a perceptually pleasing result.

 

[gollywop]

I'm beginning to realize that this dtmateojr guy (Demosthenes Mateo Jr.) is simply in seventh heaven playing in the sandboxes he creates here on dpr. After observing his behavior in a couple of recent threads, I became curious to figure out what might be motivating it, and I believe that it that he is just plainly overjoyed at having finally found a way to get people to talk to him. He needs attention at any cost.

It is of interest to note, for example, that his twitter page, which began in May 2011, has 69 tweets and absolutely zero replies over two years. His last tweet was July 6, 2013, shortly after his joining dpr on June 13, 2013. In this new venue he has clearly found a way to attract attention, and as long as he continues to get it, I am convinced he will just go on and on with his current purposely inane and obtuse behavior.

Bear this in mind when you consider responding to his posts. He isn't going to change. He has no reason to. He has what he wants/needs, and you will just be feeding its continuation.

You will also note that the comments made to his famously fatuous post, to which he frequently refers, are best summed up with reilly diefenbach's lead-off comment: Possibly the silliest post having to do with photography anyone has ever made.

He is apparently also (are you ready for this?) a RedHat Certified Engineer, an accomplishment for which he is inordinately proud. You can glean another dimension of his personality from his response to the question as to what being a RHCE means to him. You will note that all the others on the page give responses that actually answer that question. dtmateojr's response, however, is nothing but a self-serving brag session that indicates that the only meaning for him of being a RHCE is in its providing an exam or two on which he could get perfect scores.

Play in his sandbox if you will, but recognize what's buried there.

--
gollywop

I think people have a right to their own persona here on DPR, even if it is a fictional one. He doesn't force people to respond to him, and this place gets very quiet without and entertainer or two about.

You're, of course, correct Bob. But sometimes the pleasure of scratching an itch turns to irritation and torment.

--
gollywop

 

[Simon Garrett]

But sometimes the pleasure of scratching an itch turns to irritation and torment.

True, but it also gives us all endless amusement.

There were a few lengthy threads on Lula (e.g. http://www.luminous-landscape.com/forum/index.php?topic=92767.0) around posts and videos by Gary Fong and Will Crocket trying to explain why Adobe RGB is worse than sRGB, and showing what happens when you convert from Adobe RGB to sRGB:


From

This clearly shows that the you lose red and violet when you convert an image from Adobe RGB to sRGB. It's true! Gary Fong is a professional protographer and has shot more than 1,000 weddings, and therefore knows more than colour scientists and people like Andrew Rodney (digitaldog).

(From

starting about 30 seconds in.)

You can't change closed minds.

PS - I've just read some of dtmateojr's blog. The clue that you're not going to read anything informative in a blog is when you see that the posts are mainly about ridiculing others, and by implication showing how clever the poster is.

--

Simon

 

[bobn2]

I'm beginning to realize that this dtmateojr guy (Demosthenes Mateo Jr.) is simply in seventh heaven playing in the sandboxes he creates here on dpr. After observing his behavior in a couple of recent threads, I became curious to figure out what might be motivating it, and I believe that it that he is just plainly overjoyed at having finally found a way to get people to talk to him. He needs attention at any cost.

It is of interest to note, for example, that his twitter page, which began in May 2011, has 69 tweets and absolutely zero replies over two years. His last tweet was July 6, 2013, shortly after his joining dpr on June 13, 2013. In this new venue he has clearly found a way to attract attention, and as long as he continues to get it, I am convinced he will just go on and on with his current purposely inane and obtuse behavior.

Bear this in mind when you consider responding to his posts. He isn't going to change. He has no reason to. He has what he wants/needs, and you will just be feeding its continuation.

You will also note that the comments made to his famously fatuous post, to which he frequently refers, are best summed up with reilly diefenbach's lead-off comment: Possibly the silliest post having to do with photography anyone has ever made.

He is apparently also (are you ready for this?) a RedHat Certified Engineer, an accomplishment for which he is inordinately proud. You can glean another dimension of his personality from his response to the question as to what being a RHCE means to him. You will note that all the others on the page give responses that actually answer that question. dtmateojr's response, however, is nothing but a self-serving brag session that indicates that the only meaning for him of being a RHCE is in its providing an exam or two on which he could get perfect scores.

Play in his sandbox if you will, but recognize what's buried there.

 

[gollywop]

But sometimes the pleasure of scratching an itch turns to irritation and torment.

True, but it also gives all endless amusement.

There were a few lengthy threads on Lula (e.g. http://www.luminous-landscape.com/forum/index.php?topic=92767.0) around posts and videos by Gary Fong and Will Crocket trying to explain why Adobe RGB is worse than sRGB, and showing what happens when you convert from Adobe RGB to sRGB:


From

This clearly shows that the you lose red and violet when you convert an image from Adobe RGB to sRGB. It's true! Gary Fong is a professional protographer and has shot more than 1,000 weddings, and therefore knows more than colour scientists and people like Andrew Rodney (digitaldog).

(From

starting about 30 seconds in.)

You can't change closed minds.

Well, I hate to say that I missed that one, so thanks for the link. It did indeed give me a hearty giggle for the day.

In the interest of complete disclosure, however, let me note that you did misrepresent his mis-explanation by using the term "convert." He was merely toggling between inappropriate monitor/display profiles. ;-)

I was particularly impressed when he ended up with the, "I hope that helps to clear up the confusion between Adobe RGB and sRGB." Yep, sure did.

So yes, you're right: dtmateojr has some real competition for the reilly diefenbach prize. Several people seem to have gotten a perfect score.

--
gollywop

 

[gollywop]

I'm beginning to realize that this dtmateojr guy (Demosthenes Mateo Jr.) is simply in seventh heaven playing in the sandboxes he creates here on dpr. After observing his behavior in a couple of recent threads, I became curious to figure out what might be motivating it, and I believe that it that he is just plainly overjoyed at having finally found a way to get people to talk to him. He needs attention at any cost.

It is of interest to note, for example, that his twitter page, which began in May 2011, has 69 tweets and absolutely zero replies over two years. His last tweet was July 6, 2013, shortly after his joining dpr on June 13, 2013. In this new venue he has clearly found a way to attract attention, and as long as he continues to get it, I am convinced he will just go on and on with his current purposely inane and obtuse behavior.

Bear this in mind when you consider responding to his posts. He isn't going to change. He has no reason to. He has what he wants/needs, and you will just be feeding its continuation.

You will also note that the comments made to his famously fatuous post, to which he frequently refers, are best summed up with reilly diefenbach's lead-off comment: Possibly the silliest post having to do with photography anyone has ever made.

He is apparently also (are you ready for this?) a RedHat Certified Engineer, an accomplishment for which he is inordinately proud. You can glean another dimension of his personality from his response to the question as to what being a RHCE means to him. You will note that all the others on the page give responses that actually answer that question. dtmateojr's response, however, is nothing but a self-serving brag session that indicates that the only meaning for him of being a RHCE is in its providing an exam or two on which he could get perfect scores.

Play in his sandbox if you will, but recognize what's buried there.

--
gollywop

I think people have a right to their own persona here on DPR, even if it is a fictional one. He doesn't force people to respond to him, and this place gets very quiet without and entertainer or two about.

You're, of course, correct Bob. But sometimes the pleasure of scratching an itch turns to irritation and torment.

That's right, you need the self discipline not to scratch. As my mother told me, if you scratch it, it will get worse.

Yeah, yeah Bob; she's right, and clearly a good mom.

But that was my point.

--
gollywop

 

[xpatUSA]

I'm beginning to realize that this dtmateojr guy (Demosthenes Mateo Jr.) is simply in seventh heaven playing in the sandboxes he creates here on dpr. After observing his behavior in a couple of recent threads, I became curious to figure out what might be motivating it, and I believe that it that he is just plainly overjoyed at having finally found a way to get people to talk to him. He needs attention at any cost.

It is of interest to note, for example, that his twitter page, which began in May 2011, has 69 tweets and absolutely zero replies over two years. His last tweet was July 6, 2013, shortly after his joining dpr on June 13, 2013. In this new venue he has clearly found a way to attract attention, and as long as he continues to get it, I am convinced he will just go on and on with his current purposely inane and obtuse behavior.

Bear this in mind when you consider responding to his posts. He isn't going to change. He has no reason to. He has what he wants/needs, and you will just be feeding its continuation.

You will also note that the comments made to his famously fatuous post, to which he frequently refers, are best summed up with reilly diefenbach's lead-off comment: Possibly the silliest post having to do with photography anyone has ever made.

He is apparently also (are you ready for this?) a RedHat Certified Engineer, an accomplishment for which he is inordinately proud. You can glean another dimension of his personality from his response to the question as to what being a RHCE means to him. You will note that all the others on the page give responses that actually answer that question. dtmateojr's response, however, is nothing but a self-serving brag session that indicates that the only meaning for him of being a RHCE is in its providing an exam or two on which he could get perfect scores.

Play in his sandbox if you will, but recognize what's buried there.

An excellent exposé!

And the alliteration in "famously fatuous post" was a joy to behold.

Well said, Sir.

 

[xpatUSA]

Gary Fong is a professional protographer and has shot more than 1,000 weddings, and therefore knows more than colour scientists and people like Andrew Rodney (digitaldog).

You struck a nerve there, Simon! I've clashed with the Dog a time or too . . .

 

[Simon Garrett]

Gary Fong is a professional protographer and has shot more than 1,000 weddings, and therefore knows more than colour scientists and people like Andrew Rodney (digitaldog).

You struck a nerve there, Simon! I've clashed with the Dog a time or too . . .

I've not always agreed with him either, but in a disagreement between him and Gary Fong about colour, I know who is more likely to be right. :-D

 

[wchutt]

I asked this before, never got an answer. How, exactly, does having more photons 'reduce the error in measurement'? It would be a good start to explain first what you think it is being measured.

 

[bobn2]

I asked this before, never got an answer. How, exactly, does having more photons 'reduce the error in measurement'? It would be a good start to explain first what you think it is being measured.

 

[xpatUSA]

Gary Fong is a professional protographer and has shot more than 1,000 weddings, and therefore knows more than colour scientists and people like Andrew Rodney (digitaldog).

You struck a nerve there, Simon! I've clashed with the Dog a time or too . . .

I've not always agreed with him either, but in a disagreement between him and Gary Fong about colour, I know who is more likely to be right. :-D

Indeed, and looking at that image in your post, I now see what you mean about Mr Fong (it didn't hit till just now, duh).

As to the Dog, it wasn't really what he said, it was how it got said. Makes one's hackles rise, in the canine sense, every time, grrrrr ;-)

--
Cheers,
Ted