Quantitative comparison of Foveon and Bayer spectral response

Started Nov 15, 2018 | Discussions
Scottelly
Scottelly Forum Pro • Posts: 14,672
Re: Checking the consistency of my results

mike earussi wrote:

richard stone wrote:

mike earussi wrote:

DavidWright2010 wrote:

FWIW, I repeated my experiment today. The variable I couldn't control was the sunlight, and there is still a lot of smoke in the air - the sky overhead is a grey-blue - nowhere near a normal deep blue.

I also adjusted the RGB levels individually to get the region outside of the spectrum as near black as possible (so that the plots would go to zero.

I added the results for a Nikon D7000.

Nikon:

Pentax:

Sigma:

And the plots of each color:

There are differences from the last set. Maybe it's the light, maybe it's my technique.

David

Again amazing images. I'm actually surprised to see how much sharper the color transitions are in Bayer compared to Foveon. Thanks for posting.

I'm not surprised by the sharp and not so sharp separations: It's exactly what you would expect, I think, with the overlapping sensitivities to the various colors as shown on every single Foveon sensor color response chart, compared to the specific color filters of CFA. But what is surprising is that people actually see (or think they see) the differences in the images. Assuming that is what we see. And at this point we begin to run into reporting and description issues.

Clearly it is easy enough to say that the idea with any imaging system has to be, at its center, that the final image is going to be a good representation of the scene.

Not necessarily, unless that's what you're wanting. Never having worked for National Geographic I've never felt the need to adhere to their strict standards of realism. I'm more than happy to manipulate the scene as I see fit to create whatever mood I want. What I'm wanting from the raw file is a lot of information to work with and Foveon provides more of that than Bayer does.

But yet, that isn't exactly right either, if we consider the multitude of film choices, designed for one purpose or another. Velvia for portraits? For weddings? Generally no. Transparency images for weddings? Not so much.

In any event, in terms of numbers compared to what it looks like: we need both to make good decisions.

It does look like the colors shown by Foveon are slightly more comprehensive than from CFA? Is that a word we might use?

Whatever word you wish to use. Just more of them is how I put it. When I take a picture of a landscape with both Bayer and Foveon I can see more subtle shades of green and browns in the Foveon shot which makes it look more realistic to me. Bayer shots can sometimes look almost monochromatic photographing pine trees, for instance, whereas the Foveon shot shows different colors of green and blue green mixed with a little orange.

I agree Mike, and the color of green from a Bayer patterned CFA sensor camera is sometimes weird and unnatural. I've seen grass that looked fake from my Sony cameras, but I don't remember ever seeing grass that looked fake in photos from my Sigma cameras . . . though the Sigmas do indeed have their own problems with color.

-- hide signature --

Scott Barton Kennelly
http://www.bigprintphotos.com

 Scottelly's gear list:Scottelly's gear list
Sony SLT-A65 Sigma SD1 Merrill Nikon D810 Sigma sd Quattro H Sony DT 18-55mm F3.5-5.6 SAM +21 more
Scottelly
Scottelly Forum Pro • Posts: 14,672
Re: Good approach, but needs minor correction

Paul_CP wrote:

There's going to be a superposition of different wavelengths at each point as pointed out above, I agree.

The mixture of wavelengths would be dependent on several variables including the width of the incoming beam, the size and shape of the prism, the distance between the screen and the prism, the angle of the screen and more. Of course if the incoming beam is infinitesimally thin, there would be less of a problem, but then, there would presumably not be enough light for the sensor.

I think there are potentially 2 solutions as above, one is to use a lens to focus the light emerging from the prism onto a screen and then to photograph that screen using the lens on the camera and the other is to point the camera lens directly at the light emerging from the prism and focus. Possibly the second of these 2 solutions is better as it doesn't introduce an extra lens and removes the need for a screen?

I have never tried to focus the light emerging from a prism and there may be problems with that.

The prism itself needs to be of high quality too, to prevent unwanted dispersion or scattering within the glass.

Also, if a screen is used, what is it made of and can we be rely on a good spectral response? I'm not an expert but this could be an issue also,

I'm sure it is an issue. The reflective properties of the "screen" the rainbow is projected on is surely significant. I think it could be similar to the way the resulting colors in a print can look different based on the way the colors of printer ink is influenced by the paper it is printed on. Printing on one paper vs another can cause an image to have obviously different colors in it. I'm sure the same is true for the rainbow projected from sunlight by a prism onto various reflective "screens." It would be interesting to experiment with this, trying a glossy white paper, a flat white paper, a bright white satin paper, and a white chalk board. I know the first test shots I will make are going to be shot with the rainbow projected onto a white foam-core board, which I have sitting against a wall in my bedroom right now. I'll have to try some white paper and maybe a white, cotton sheet too.

-- hide signature --

Scott Barton Kennelly
http://www.bigprintphotos.com

 Scottelly's gear list:Scottelly's gear list
Sony SLT-A65 Sigma SD1 Merrill Nikon D810 Sigma sd Quattro H Sony DT 18-55mm F3.5-5.6 SAM +21 more
Scottelly
Scottelly Forum Pro • Posts: 14,672
Re: Quantitative comparison of Foveon and Bayer spectral response

xpatUSA wrote:

mike earussi wrote:

xpatUSA wrote:

mike earussi wrote:

I am amazed (though perhaps I shouldn't be) about how your straight forward and easily understood photo could degenerate into a complex discussion of numbers.

The use of "degenerate" as opposed to a less derogatory word is quite noticeable to the pedantic eye.

Not really, that's exactly what I meant. Technical discussions can be fine, but when they replace and even supersede the eye then I think you are as least putting the "cart before the horse" so to speak.

I don't know about anyone else here but I don't photograph test charts and diagrams, I prefer nature. And it's easy to see from your photos the clear difference between the two sensors. I can only speak for myself, but my eyes don't need complex graphs and charts to see the differences between your two shots.

Based on such an impassioned diatribe, it's time for me to unsubscribe ...

Ted, your choice, but are your photos based on numbers or on what you see with your eyes?

My diabetic eyes are pretty bad as would images based on what I see, especially in the world of color, unfortunately.

I use an incident light-meter and enter the exposure numbers it gives into the camera, based on my desired f/number. Based on the scene, I offset those numbers by an exposure compensation value. I open the X3F in RawDigger and review the numbers shown by the raw histogram, with a passing glance at the over/under exposure numbers in the header.

Off to SPP for conversion and export. I take notice of the color-picker numbers, especially if color-balancing, during editing. During exporting, e.g. for the web, I select a JPEG Quality number . Or for export to RawTherapee, I choose between the bit-depth numbers for TIFF.

I could go on but how your photos get produced without numbers is a little beyond me. I supposed one could chimp on the LCD and adjust one's images on-screen only by eye or use up lots of printing medium by cut-and-try ...

Numbers are useful in their place. Time for an example, with lots of spin, of course.

My turn for rhetoric: You wish to match some paint on your house; somewhere you've got the old can. Do you go to the hardware store and try to pick the color from their paint swatch by eyeor do you find that old can and take the Pantone (or whatever) NUMBERS with you? No need to answer - it's a rhetorical question.

But doing it visually is SO much more fun Ted . . . especially when the ambient light in the room changes from time to time! It's just like when the sun hides behind a cloud, when you're trying to shoot test shots with the same lens and different cameras.

-- hide signature --

Scott Barton Kennelly
http://www.bigprintphotos.com

 Scottelly's gear list:Scottelly's gear list
Sony SLT-A65 Sigma SD1 Merrill Nikon D810 Sigma sd Quattro H Sony DT 18-55mm F3.5-5.6 SAM +21 more
TN Args
TN Args Veteran Member • Posts: 9,326
Re: Quantitative comparison of Foveon and Bayer spectral response
2

mike earussi wrote:

DavidWright2010 wrote:

In a previous discussion, I was trying to design a simple, reproducible, way to put numbers on the color precision (ability to distinguish colors that are very similar) of Foveon and Bayer sensors.

....So my current test uses a prism, which I bought from Amazon for $10. I set the prism to intercept a beam of sunlight coming through a window, and display the resultant spectrum on a white surface. I then take a picture of the spectrum with each camera.

You had a very good idea photographing the entire spectrum....

To be honest, I think it was my idea, link.

I didn't realise there would be a question of focus. I thought, using direct sunlight, the incoming rays would be perfectly parallel.

Of course, most colours are not shown at all on such a projection.

 TN Args's gear list:TN Args's gear list
Sigma dp0 Quattro Panasonic Lumix DMC-GX7 Olympus E-M5 II Olympus M.Zuiko Digital ED 9-18mm F4.0-5.6 Panasonic Lumix G 14mm F2.5 ASPH +7 more
docmaas
docmaas Veteran Member • Posts: 6,821
Re: Quantitative comparison of Foveon and Bayer spectral response

COOL OFF.  The word "degenerate" means two different things: "break down" as a verb or a person of weak moral quality as an noun.   Mike pretty clearly meant "break down."  Certainly that might have been a little objectionable but not sufficiently so to break off the conversation.  Sometimes though the second less preferable meaning can still taint the reader even if it was not intended to do so.

mike earussi wrote:

xpatUSA wrote:

mike earussi wrote:

I am amazed (though perhaps I shouldn't be) about how your straight forward and easily understood photo could degenerate into a complex discussion of numbers.

The use of "degenerate" as opposed to a less derogatory word is quite noticeable to the pedantic eye.

Not really, that's exactly what I meant. Technical discussions can be fine, but when they replace and even supersede the eye then I think you are as least putting the "cart before the horse" so to speak.

I don't know about anyone else here but I don't photograph test charts and diagrams, I prefer nature. And it's easy to see from your photos the clear difference between the two sensors. I can only speak for myself, but my eyes don't need complex graphs and charts to see the differences between your two shots.

Based on such an impassioned diatribe, it's time for me to unsubscribe ...

Ted, your choice, but are your photos based on numbers or on what you see with your eyes?

-- hide signature --

What would Bates do?
"For me, photography is only an artistic language. The camera is my pencil. -- Charly Ho
"At every crossroads on the path that leads to the future, tradition has placed 10,000 men to guard the past."
Maurice Maeterlinck

xpatUSA
xpatUSA Forum Pro • Posts: 17,175
Re: Quantitative comparison of Foveon and Bayer spectral response

TN Args wrote:

mike earussi wrote:

DavidWright2010 wrote:

In a previous discussion, I was trying to design a simple, reproducible, way to put numbers on the color precision (ability to distinguish colors that are very similar) of Foveon and Bayer sensors.

....So my current test uses a prism, which I bought from Amazon for $10. I set the prism to intercept a beam of sunlight coming through a window, and display the resultant spectrum on a white surface. I then take a picture of the spectrum with each camera.

You had a very good idea photographing the entire spectrum....

To be honest, I think it was my idea, link.

I didn't realise there would be a question of focus. I thought, using direct sunlight, the incoming rays would be perfectly parallel.

Of course, most colours are not shown at all on such a projection.

Good point.

-- hide signature --

Ted

 xpatUSA's gear list:xpatUSA's gear list
Sigma DP2 Panasonic Lumix DMC-G1 Sigma SD15 Panasonic Lumix DMC-GH1 Panasonic Lumix DC-G9 +14 more
TN Args
TN Args Veteran Member • Posts: 9,326
Re: Quantitative comparison of Foveon and Bayer spectral response

It would be nice to see the raw files please.

 TN Args's gear list:TN Args's gear list
Sigma dp0 Quattro Panasonic Lumix DMC-GX7 Olympus E-M5 II Olympus M.Zuiko Digital ED 9-18mm F4.0-5.6 Panasonic Lumix G 14mm F2.5 ASPH +7 more
TN Args
TN Args Veteran Member • Posts: 9,326
Re: Quantitative comparison of Foveon and Bayer spectral response
2

docmaas wrote:

COOL OFF. The word "degenerate" means two different things: "break down" as a verb or a person of weak moral quality as an noun. Mike pretty clearly meant "break down." Certainly that might have been a little objectionable but not sufficiently so to break off the conversation. Sometimes though the second less preferable meaning can still taint the reader even if it was not intended to do so.

mike earussi wrote:

xpatUSA wrote:

mike earussi wrote:

I am amazed (though perhaps I shouldn't be) about how your straight forward and easily understood photo could degenerate into a complex discussion of numbers.

Perhaps, Mike, you can check the thread-starting post and the purpose of this thread.

Far from 'degenerating' the thread, people are trying to fulfil its purpose.

The use of "degenerate" as opposed to a less derogatory word is quite noticeable to the pedantic eye.

Not really, that's exactly what I meant. Technical discussions can be fine, but when they replace and even supersede the eye then I think you are as least putting the "cart before the horse" so to speak.

Even the title of the thread should be a giveaway.

I don't know about anyone else here but I don't photograph test charts and diagrams, I prefer nature. And it's easy to see from your photos the clear difference between the two sensors. I can only speak for myself, but my eyes don't need complex graphs and charts to see the differences between your two shots.

Based on such an impassioned diatribe, it's time for me to unsubscribe ...

Ted, your choice, but are your photos based on numbers or on what you see with your eyes?

TITLE?

 TN Args's gear list:TN Args's gear list
Sigma dp0 Quattro Panasonic Lumix DMC-GX7 Olympus E-M5 II Olympus M.Zuiko Digital ED 9-18mm F4.0-5.6 Panasonic Lumix G 14mm F2.5 ASPH +7 more
docmaas
docmaas Veteran Member • Posts: 6,821
Re: Quantitative comparison of Foveon and Bayer spectral response

Too many Mikes!  Your response is to my post but deals entirely with Mike Earussi's text.

Mike

TN Args wrote:

docmaas wrote:

COOL OFF. The word "degenerate" means two different things: "break down" as a verb or a person of weak moral quality as an noun. Mike pretty clearly meant "break down." Certainly that might have been a little objectionable but not sufficiently so to break off the conversation. Sometimes though the second less preferable meaning can still taint the reader even if it was not intended to do so.

mike earussi wrote:

xpatUSA wrote:

mike earussi wrote:

I am amazed (though perhaps I shouldn't be) about how your straight forward and easily understood photo could degenerate into a complex discussion of numbers.

Perhaps, Mike, you can check the thread-starting post and the purpose of this thread.

Far from 'degenerating' the thread, people are trying to fulfil its purpose.

The use of "degenerate" as opposed to a less derogatory word is quite noticeable to the pedantic eye.

Not really, that's exactly what I meant. Technical discussions can be fine, but when they replace and even supersede the eye then I think you are as least putting the "cart before the horse" so to speak.

Even the title of the thread should be a giveaway.

I don't know about anyone else here but I don't photograph test charts and diagrams, I prefer nature. And it's easy to see from your photos the clear difference between the two sensors. I can only speak for myself, but my eyes don't need complex graphs and charts to see the differences between your two shots.

Based on such an impassioned diatribe, it's time for me to unsubscribe ...

Ted, your choice, but are your photos based on numbers or on what you see with your eyes?

TITLE?

-- hide signature --

What would Bates do?
"For me, photography is only an artistic language. The camera is my pencil. -- Charly Ho
"At every crossroads on the path that leads to the future, tradition has placed 10,000 men to guard the past."
Maurice Maeterlinck

TN Args
TN Args Veteran Member • Posts: 9,326
Re: Quantitative comparison of Foveon and Bayer spectral response

Sorry.

 TN Args's gear list:TN Args's gear list
Sigma dp0 Quattro Panasonic Lumix DMC-GX7 Olympus E-M5 II Olympus M.Zuiko Digital ED 9-18mm F4.0-5.6 Panasonic Lumix G 14mm F2.5 ASPH +7 more
OP DavidWright2010 Senior Member • Posts: 2,169
Re: Quantitative comparison of Foveon and Bayer spectral response
1

TN Args wrote:

It would be nice to see the raw files please.

This is the second set.

Let's see if that works.

David

 DavidWright2010's gear list:DavidWright2010's gear list
Sigma DP1 Merrill Sigma DP2 Merrill Sigma DP3 Merrill Sigma SD1 Merrill Pentax K-1 +1 more
xpatUSA
xpatUSA Forum Pro • Posts: 17,175
Re: Quantitative comparison of Foveon and Bayer spectral response

DavidWright2010 wrote:

TN Args wrote:

It would be nice to see the raw files please.

This is the second set.

Let's see if that works.

Thanks for the raws, David. I looked at the X3F in RawDigger:

A lot of underexposed area but the spectrum area is not showing warnings.

Then I exported the top layer direct to ImageJ:

Not so hot, I'm afraid. With that much noise, gradation is quite hard to assess, even just with one's eyes.

Then I exported the RawDigger RGB image to ImageJ and split the colors. Here is the blue "channel", post RD's conversion:

Bit of a joke really. This is what SPP and Kalpanika have to put up with - a lesson for those who like to shoot their Merrills at 200+ ISO ...

"Waiter!" <snap, snap> "I need more noise in my Merrill image!"

"Certainly Sir!" <adjusts ISO knob to 400> ...

-- hide signature --

Ted

 xpatUSA's gear list:xpatUSA's gear list
Sigma DP2 Panasonic Lumix DMC-G1 Sigma SD15 Panasonic Lumix DMC-GH1 Panasonic Lumix DC-G9 +14 more
DMillier Forum Pro • Posts: 21,612
How science works...
1

mike earussi wrote:

TN Args wrote:

mike earussi wrote:

DavidWright2010 wrote:

mike earussi wrote:

You had a very good idea photographing the entire spectrum. This demonstrates perfectly the point I and others have been trying to make about the superiority of the color response of the Foveon to the Bayer. The DP3M shows a very smooth transition from one color to the next whereas the Bayer is more abrupt. The photographs remind me of the differences between 8 bit color vs. 16 bit color.

Yes, my results (if correct) show a sharper descent of the Bayer filter transmission.

Bayer's fixed color filters do not have the ability to show as many colors as the Foveon chip does and they never will.

You may be right. I, too, was unable to replicate the varying hues of a distant green field (recorded with a Foveon sensor) with a Bayer sensor. I'm just trying to put numbers on the effect.

P.S. You did this with the Merrill chip. It would also be interesting to do this with a Quattro chip to see if there is any difference.

I don't have a Quattro; maybe someone else would like to try. I bought this $11 prism from Amazon . And my setup is simple:

David

I am amazed (though perhaps I shouldn't be) about how your straight forward and easily understood photo could degenerate into a complex discussion of numbers. I don't know about anyone else here but I don't photograph test charts and diagrams, I prefer nature. And it's easy to see from your photos the clear difference between the two sensors. I can only speak for myself, but my eyes don't need complex graphs and charts to see the differences between your two shots.

So again I thank you for a very clear and easily understood visual demonstration of the Bayer vs. Foveon differences.

However, your interpretation was pretty wild, and the numerical analysis is trying to take that bias out of it.

My interpretation is the only thing that matters to me, just like everyone else has their personal taste and interpretation. I prefer what I see in the Foveon compared to what I see in the Bayer. If you wish or feel the need to analyze it or prefer the Bayer to the Foveon that's your choice. But the impression I've got from these discussions is that some don't trust their eyes or refuse to believe what they see without numbers to back it up. I don't require that. My eyes are all I need to form a judgment. But if you need numbers have fun.

I'm not going to argue with your personal preferences, they are yours and yours alone. However, I want to point out something to defend one reason why people attempt tests like this and why (if done well) they are useful:   confirmation bias.

It is quite natural for people to reach for evidence - any evidence - that supports their view and disregard contrary evidence.  I'm not insulting anyone by pointing this out, it's the way brains work and everybody - everybody - is subject to this bias (among many others). For example, it is clear as day in your post above. You looked at the shots and made your mind up instantly that they support your belief that Foveon is better.

The problem here is ascertaining whether this is what experiment really shows us.

To do that, the experiment has to be well designed, all the errors and biases in it have to be quantified and understood until it has been "purified" to the point where we can be confident that the results its appears to give are actually correct (rather than an artefact of the experiment).

Your response above shows that you are not prepared to wait for that process because you can see something in the initial results that appears to support your beliefs and you want to cut things at this point and prevent any further investigation to avoid any possibility of contrary evidence emerging.

This is not a criticism of you alone, it applies to everyone, but it is a good example of why science has had to develop protocols to try and minimise the natural biases in all of us. As Feynman said, the easiest person to fool is yourself.  Reducing bias is the only chance we have of getting anywhere near to the truth (which often tells us things we hate hearing).

IMO, at this point in this experiment/debate, nothing either way had been demonstrated. We are collectively groping towards refining an experiment in the hope it will eventually reliably show us something trustworthy.

Turning things into numbers is part of the process of objectification.

Whether you accept or deny the necessity of these steps is maybe revealing of openness of mind and self reflection but not support either way for one technology or another.

-- hide signature --
DMillier Forum Pro • Posts: 21,612
Re: Checking the consistency of my results
2

richard stone wrote:

mike earussi wrote:

DavidWright2010 wrote:

FWIW, I repeated my experiment today. The variable I couldn't control was the sunlight, and there is still a lot of smoke in the air - the sky overhead is a grey-blue - nowhere near a normal deep blue.

I also adjusted the RGB levels individually to get the region outside of the spectrum as near black as possible (so that the plots would go to zero.

I added the results for a Nikon D7000.

Nikon:

Pentax:

Sigma:

And the plots of each color:

There are differences from the last set. Maybe it's the light, maybe it's my technique.

David

Again amazing images. I'm actually surprised to see how much sharper the color transitions are in Bayer compared to Foveon. Thanks for posting.

I'm not surprised by the sharp and not so sharp separations: It's exactly what you would expect, I think, with the overlapping sensitivities to the various colors as shown on every single Foveon sensor color response chart, compared to the specific color filters of CFA. But what is surprising is that people actually see (or think they see) the differences in the images. Assuming that is what we see. And at this point we begin to run into reporting and description issues.

Clearly it is easy enough to say that the idea with any imaging system has to be, at its center, that the final image is going to be a good representation of the scene.

But yet, that isn't exactly right either, if we consider the multitude of film choices, designed for one purpose or another. Velvia for portraits? For weddings? Generally no. Transparency images for weddings? Not so much.

In any event, in terms of numbers compared to what it looks like: we need both to make good decisions.

It does look like the colors shown by Foveon are slightly more comprehensive than from CFA? Is that a word we might use?

Some points:

1. It is well understood I believe that the choice of filters for Bayer CFA is made to be a good match for the eye's response.

2. Good but not perfect - Phase One have their special Trichromatic filters where they have gone for as perfect a match as possible even if it compromises other photographic parameters such as noise/dynamic range.  Those people are not fools, I think you need to give them some credit for understanding colour.

3. It is also well know that Foveon sensors filtration through silicon depth penetration produces weird results that need transformation into human colour space.

These starting facts make me sceptical that Foveon colour is likely to be as good as the most optimised Bayer sensors like the special Phase One designs.  You really need to have some kind of justification as why it is likely to before honestly being able for hold the premise that Foveon should be better. It is a much more likely hypothesis that Bayer is a good approximation to human colour vision while you might expect some weirdnesses in Foveon results.  And certainly that is my experience over several Foveon models:  I sometimes get weird looking colour from Foveon and overall I prefer Bayer colour because it rarely jars.

Now given that piece of theorising, we should be sceptical of any test results that go against that without being completely sure we have ironed all the wrinkles out of the experimental method and its analysis.

if after accounting for experimental artefacts, we come to a trustworthy conclusion that Foveon colour response is superior to Bayer, that would be a big surprise to me but if confirmed, would have to be accepted.

But at this point in the proceedings, given what I already think about Foveon colour (my starting biases if you will), I remain sceptical about this experiment. It seems to me that as amateur scientists, the experiment is bound to have flaws - whatever it purports to show.

However, it is good to try!

-- hide signature --
DMillier Forum Pro • Posts: 21,612
Re: Checking the consistency of my results
1

Scottelly wrote:

mike earussi wrote:

richard stone wrote:

mike earussi wrote:

DavidWright2010 wrote:

FWIW, I repeated my experiment today. The variable I couldn't control was the sunlight, and there is still a lot of smoke in the air - the sky overhead is a grey-blue - nowhere near a normal deep blue.

I also adjusted the RGB levels individually to get the region outside of the spectrum as near black as possible (so that the plots would go to zero.

I added the results for a Nikon D7000.

Nikon:

Pentax:

Sigma:

And the plots of each color:

There are differences from the last set. Maybe it's the light, maybe it's my technique.

David

Again amazing images. I'm actually surprised to see how much sharper the color transitions are in Bayer compared to Foveon. Thanks for posting.

I'm not surprised by the sharp and not so sharp separations: It's exactly what you would expect, I think, with the overlapping sensitivities to the various colors as shown on every single Foveon sensor color response chart, compared to the specific color filters of CFA. But what is surprising is that people actually see (or think they see) the differences in the images. Assuming that is what we see. And at this point we begin to run into reporting and description issues.

Clearly it is easy enough to say that the idea with any imaging system has to be, at its center, that the final image is going to be a good representation of the scene.

Not necessarily, unless that's what you're wanting. Never having worked for National Geographic I've never felt the need to adhere to their strict standards of realism. I'm more than happy to manipulate the scene as I see fit to create whatever mood I want. What I'm wanting from the raw file is a lot of information to work with and Foveon provides more of that than Bayer does.

But yet, that isn't exactly right either, if we consider the multitude of film choices, designed for one purpose or another. Velvia for portraits? For weddings? Generally no. Transparency images for weddings? Not so much.

In any event, in terms of numbers compared to what it looks like: we need both to make good decisions.

It does look like the colors shown by Foveon are slightly more comprehensive than from CFA? Is that a word we might use?

Whatever word you wish to use. Just more of them is how I put it. When I take a picture of a landscape with both Bayer and Foveon I can see more subtle shades of green and browns in the Foveon shot which makes it look more realistic to me. Bayer shots can sometimes look almost monochromatic photographing pine trees, for instance, whereas the Foveon shot shows different colors of green and blue green mixed with a little orange.

I agree Mike, and the color of green from a Bayer patterned CFA sensor camera is sometimes weird and unnatural. I've seen grass that looked fake from my Sony cameras, but I don't remember ever seeing grass that looked fake in photos from my Sigma cameras . . . though the Sigmas do indeed have their own problems with color.

Have you never looked at grass and yellowed grass shot with a SD9 or SD14 for example?

They can so some horrible things, especially under morning or evening light.

The SD9 is also a disaster with greens. It often produced a greyed, low saturation effect (see Phil Askey's original SD9 review for examples).

When my daughter was born I was in the middle of my sD9 period. Looking at the pictures from the SD9, her pink skin often looks jaundiced. My Bayer shots didn't do that.

There is quite a mix of things going on here. Colour in photography is rarely tuned to be accurate but rather to be "nice". This fact can skew our ideas of what camera is doing what with colour.  And achieving a nice effect in one area of the spectrum can often cause problems in other areas - whatever the technology.

-- hide signature --
TN Args
TN Args Veteran Member • Posts: 9,326
Re: How science works...

DMillier wrote:

mike earussi wrote:

TN Args wrote:

mike earussi wrote:

DavidWright2010 wrote:

mike earussi wrote:

You had a very good idea photographing the entire spectrum. This demonstrates perfectly the point I and others have been trying to make about the superiority of the color response of the Foveon to the Bayer. The DP3M shows a very smooth transition from one color to the next whereas the Bayer is more abrupt. The photographs remind me of the differences between 8 bit color vs. 16 bit color.

Yes, my results (if correct) show a sharper descent of the Bayer filter transmission.

Bayer's fixed color filters do not have the ability to show as many colors as the Foveon chip does and they never will.

You may be right. I, too, was unable to replicate the varying hues of a distant green field (recorded with a Foveon sensor) with a Bayer sensor. I'm just trying to put numbers on the effect.

P.S. You did this with the Merrill chip. It would also be interesting to do this with a Quattro chip to see if there is any difference.

I don't have a Quattro; maybe someone else would like to try. I bought this $11 prism from Amazon . And my setup is simple:

David

I am amazed (though perhaps I shouldn't be) about how your straight forward and easily understood photo could degenerate into a complex discussion of numbers. I don't know about anyone else here but I don't photograph test charts and diagrams, I prefer nature. And it's easy to see from your photos the clear difference between the two sensors. I can only speak for myself, but my eyes don't need complex graphs and charts to see the differences between your two shots.

So again I thank you for a very clear and easily understood visual demonstration of the Bayer vs. Foveon differences.

However, your interpretation was pretty wild, and the numerical analysis is trying to take that bias out of it.

My interpretation is the only thing that matters to me, just like everyone else has their personal taste and interpretation. I prefer what I see in the Foveon compared to what I see in the Bayer. If you wish or feel the need to analyze it or prefer the Bayer to the Foveon that's your choice. But the impression I've got from these discussions is that some don't trust their eyes or refuse to believe what they see without numbers to back it up. I don't require that. My eyes are all I need to form a judgment. But if you need numbers have fun.

I'm not going to argue with your personal preferences, they are yours and yours alone. However, I want to point out something to defend one reason why people attempt tests like this and why (if done well) they are useful: confirmation bias.

It is quite natural for people to reach for evidence - any evidence - that supports their view and disregard contrary evidence. I'm not insulting anyone by pointing this out, it's the way brains work and everybody - everybody - is subject to this bias (among many others). For example, it is clear as day in your post above. You looked at the shots and made your mind up instantly that they support your belief that Foveon is better.

The problem here is ascertaining whether this is what experiment really shows us.

To do that, the experiment has to be well designed, all the errors and biases in it have to be quantified and understood until it has been "purified" to the point where we can be confident that the results its appears to give are actually correct (rather than an artefact of the experiment).

Your response above shows that you are not prepared to wait for that process because you can see something in the initial results that appears to support your beliefs and you want to cut things at this point and prevent any further investigation to avoid any possibility of contrary evidence emerging.

This is not a criticism of you alone, it applies to everyone, but it is a good example of why science has had to develop protocols to try and minimise the natural biases in all of us. As Feynman said, the easiest person to fool is yourself. Reducing bias is the only chance we have of getting anywhere near to the truth (which often tells us things we hate hearing).

IMO, at this point in this experiment/debate, nothing either way had been demonstrated. We are collectively groping towards refining an experiment in the hope it will eventually reliably show us something trustworthy.

Turning things into numbers is part of the process of objectification.

Whether you accept or deny the necessity of these steps is maybe revealing of openness of mind and self reflection but not support either way for one technology or another.

Yes, that's the long version of my 1-liner, above.  

 TN Args's gear list:TN Args's gear list
Sigma dp0 Quattro Panasonic Lumix DMC-GX7 Olympus E-M5 II Olympus M.Zuiko Digital ED 9-18mm F4.0-5.6 Panasonic Lumix G 14mm F2.5 ASPH +7 more
TN Args
TN Args Veteran Member • Posts: 9,326
Re: Checking the consistency of my results

DMillier wrote:

richard stone wrote:

mike earussi wrote:

DavidWright2010 wrote:

FWIW, I repeated my experiment today. The variable I couldn't control was the sunlight, and there is still a lot of smoke in the air - the sky overhead is a grey-blue - nowhere near a normal deep blue.

I also adjusted the RGB levels individually to get the region outside of the spectrum as near black as possible (so that the plots would go to zero.

I added the results for a Nikon D7000.

Nikon:

Pentax:

Sigma:

And the plots of each color:

There are differences from the last set. Maybe it's the light, maybe it's my technique.

David

Again amazing images. I'm actually surprised to see how much sharper the color transitions are in Bayer compared to Foveon. Thanks for posting.

I'm not surprised by the sharp and not so sharp separations: It's exactly what you would expect, I think, with the overlapping sensitivities to the various colors as shown on every single Foveon sensor color response chart, compared to the specific color filters of CFA. But what is surprising is that people actually see (or think they see) the differences in the images. Assuming that is what we see. And at this point we begin to run into reporting and description issues.

Clearly it is easy enough to say that the idea with any imaging system has to be, at its center, that the final image is going to be a good representation of the scene.

But yet, that isn't exactly right either, if we consider the multitude of film choices, designed for one purpose or another. Velvia for portraits? For weddings? Generally no. Transparency images for weddings? Not so much.

In any event, in terms of numbers compared to what it looks like: we need both to make good decisions.

It does look like the colors shown by Foveon are slightly more comprehensive than from CFA? Is that a word we might use?

Some points:

1. It is well understood I believe that the choice of filters for Bayer CFA is made to be a good match for the eye's response.

Well understood by me, anyway.

2. Good but not perfect - Phase One have their special Trichromatic filters where they have gone for as perfect a match as possible even if it compromises other photographic parameters such as noise/dynamic range. Those people are not fools, I think you need to give them some credit for understanding colour.

Absolutely.

3. It is also well know that Foveon sensors filtration through silicon depth penetration produces weird results that need transformation into human colour space.

A nice explanation is here, link.

These starting facts make me sceptical that Foveon colour is likely to be as good as the most optimised Bayer sensors like the special Phase One designs. You really need to have some kind of justification as why it is likely to before honestly being able for hold the premise that Foveon should be better. It is a much more likely hypothesis that Bayer is a good approximation to human colour vision while you might expect some weirdnesses in Foveon results. And certainly that is my experience over several Foveon models: I sometimes get weird looking colour from Foveon and overall I prefer Bayer colour because it rarely jars.

My attempt to explain that is here, link.

As I said there, noise is a major enemy of smooth colour gradation. And Ted has shown, a few posts above, that this case is no exception.

Stepping  back into theory for a moment, given that Foveon has larger factors in the colour transformation matrix (due to the less human-like set of filters), then quantum changes in the raw values are going to be multiplied larger into larger increments in RGB colour, i.e. bigger step gradations. But this effect is most likely drowned in noise.

cheers

 TN Args's gear list:TN Args's gear list
Sigma dp0 Quattro Panasonic Lumix DMC-GX7 Olympus E-M5 II Olympus M.Zuiko Digital ED 9-18mm F4.0-5.6 Panasonic Lumix G 14mm F2.5 ASPH +7 more
DMillier Forum Pro • Posts: 21,612
Re: How science works...

TN Args wrote:

DMillier wrote:

mike earussi wrote:

TN Args wrote:

mike earussi wrote:

DavidWright2010 wrote:

mike earussi wrote:

You had a very good idea photographing the entire spectrum. This demonstrates perfectly the point I and others have been trying to make about the superiority of the color response of the Foveon to the Bayer. The DP3M shows a very smooth transition from one color to the next whereas the Bayer is more abrupt. The photographs remind me of the differences between 8 bit color vs. 16 bit color.

Yes, my results (if correct) show a sharper descent of the Bayer filter transmission.

Bayer's fixed color filters do not have the ability to show as many colors as the Foveon chip does and they never will.

You may be right. I, too, was unable to replicate the varying hues of a distant green field (recorded with a Foveon sensor) with a Bayer sensor. I'm just trying to put numbers on the effect.

P.S. You did this with the Merrill chip. It would also be interesting to do this with a Quattro chip to see if there is any difference.

I don't have a Quattro; maybe someone else would like to try. I bought this $11 prism from Amazon . And my setup is simple:

David

I am amazed (though perhaps I shouldn't be) about how your straight forward and easily understood photo could degenerate into a complex discussion of numbers. I don't know about anyone else here but I don't photograph test charts and diagrams, I prefer nature. And it's easy to see from your photos the clear difference between the two sensors. I can only speak for myself, but my eyes don't need complex graphs and charts to see the differences between your two shots.

So again I thank you for a very clear and easily understood visual demonstration of the Bayer vs. Foveon differences.

However, your interpretation was pretty wild, and the numerical analysis is trying to take that bias out of it.

My interpretation is the only thing that matters to me, just like everyone else has their personal taste and interpretation. I prefer what I see in the Foveon compared to what I see in the Bayer. If you wish or feel the need to analyze it or prefer the Bayer to the Foveon that's your choice. But the impression I've got from these discussions is that some don't trust their eyes or refuse to believe what they see without numbers to back it up. I don't require that. My eyes are all I need to form a judgment. But if you need numbers have fun.

I'm not going to argue with your personal preferences, they are yours and yours alone. However, I want to point out something to defend one reason why people attempt tests like this and why (if done well) they are useful: confirmation bias.

It is quite natural for people to reach for evidence - any evidence - that supports their view and disregard contrary evidence. I'm not insulting anyone by pointing this out, it's the way brains work and everybody - everybody - is subject to this bias (among many others). For example, it is clear as day in your post above. You looked at the shots and made your mind up instantly that they support your belief that Foveon is better.

The problem here is ascertaining whether this is what experiment really shows us.

To do that, the experiment has to be well designed, all the errors and biases in it have to be quantified and understood until it has been "purified" to the point where we can be confident that the results its appears to give are actually correct (rather than an artefact of the experiment).

Your response above shows that you are not prepared to wait for that process because you can see something in the initial results that appears to support your beliefs and you want to cut things at this point and prevent any further investigation to avoid any possibility of contrary evidence emerging.

This is not a criticism of you alone, it applies to everyone, but it is a good example of why science has had to develop protocols to try and minimise the natural biases in all of us. As Feynman said, the easiest person to fool is yourself. Reducing bias is the only chance we have of getting anywhere near to the truth (which often tells us things we hate hearing).

IMO, at this point in this experiment/debate, nothing either way had been demonstrated. We are collectively groping towards refining an experiment in the hope it will eventually reliably show us something trustworthy.

Turning things into numbers is part of the process of objectification.

Whether you accept or deny the necessity of these steps is maybe revealing of openness of mind and self reflection but not support either way for one technology or another.

Yes, that's the long version of my 1-liner, above.

Sorry, didn't have time to make it shorter...

-- hide signature --
docmaas
docmaas Veteran Member • Posts: 6,821
Re: How science works...

Hear Hear!  Well written Dave.

My interpretation is the only thing that matters to me, just like everyone else has their personal taste and interpretation. I prefer what I see in the Foveon compared to what I see in the Bayer. If you wish or feel the need to analyze it or prefer the Bayer to the Foveon that's your choice. But the impression I've got from these discussions is that some don't trust their eyes or refuse to believe what they see without numbers to back it up. I don't require that. My eyes are all I need to form a judgment. But if you need numbers have fun.

I'm not going to argue with your personal preferences, they are yours and yours alone. However, I want to point out something to defend one reason why people attempt tests like this and why (if done well) they are useful: confirmation bias.

It is quite natural for people to reach for evidence - any evidence - that supports their view and disregard contrary evidence. I'm not insulting anyone by pointing this out, it's the way brains work and everybody - everybody - is subject to this bias (among many others). For example, it is clear as day in your post above. You looked at the shots and made your mind up instantly that they support your belief that Foveon is better.

The problem here is ascertaining whether this is what experiment really shows us.

To do that, the experiment has to be well designed, all the errors and biases in it have to be quantified and understood until it has been "purified" to the point where we can be confident that the results its appears to give are actually correct (rather than an artefact of the experiment).

Your response above shows that you are not prepared to wait for that process because you can see something in the initial results that appears to support your beliefs and you want to cut things at this point and prevent any further investigation to avoid any possibility of contrary evidence emerging.

This is not a criticism of you alone, it applies to everyone, but it is a good example of why science has had to develop protocols to try and minimise the natural biases in all of us. As Feynman said, the easiest person to fool is yourself. Reducing bias is the only chance we have of getting anywhere near to the truth (which often tells us things we hate hearing).

IMO, at this point in this experiment/debate, nothing either way had been demonstrated. We are collectively groping towards refining an experiment in the hope it will eventually reliably show us something trustworthy.

Turning things into numbers is part of the process of objectification.

Whether you accept or deny the necessity of these steps is maybe revealing of openness of mind and self reflection but not support either way for one technology or another.

-- hide signature --

What would Bates do?
"For me, photography is only an artistic language. The camera is my pencil. -- Charly Ho
"At every crossroads on the path that leads to the future, tradition has placed 10,000 men to guard the past."
Maurice Maeterlinck

DMillier Forum Pro • Posts: 21,612
Re: How science works...
1

Thanks for the kind words, Mike, although I make no claims for original thinking.

I'm in the middle of reading theoretical physicist Sabine Hossenfelder's book "Lost in Math: How beauty leads physics astray" for an interesting take on how not dealing with human biases can drive a whole community of experts to turn into alchemists.

Did you know (I didn't) that particles physics has not confirmed any predictions made post 1973? The recent Higgs confirmation at LHC was a prediction made in the 1960s.

The LHC has seen no trace of any post standard model physics and nothing supporting the string theory ideas that have dominated the community for the last 40 years.  It won't be the end of string theory as the community has become so insular they have got very good at protecting their employment prospects come what may, but it seems to be becoming clearer and clearer that the last two generations of foundational physicists have been barking up the wrong forest.

-- hide signature --
Keyboard shortcuts:
FForum MMy threads