Reading mode: Light Dark

Bayer close to limit, Foveon just started?

Go to last post
Kendall Helmstetter Gelner said:
You may be tired of that one. How about the M9 then? Same issue.

Click to expand...
The railings are clearly defined in front of the runner and if there's a colour shift in the railings in front of the pink box, my eyes are hurting too much from the colour clash to see it! And what has lens CA got to do with the sensor?

Your images are also very blurred. Either they were out of focus, or you up-interpolated them. It is a matter of fact that if you go past the resolution limit of the sensor, colours will blur together, no matter what sensor technology you use.

This is from the DP1:



See all the orange lines turning purple--even for the clearly defined semicircular lines? Oh what lousy colour resolution--from a Foveon sensor! Oh by the way, that shouldn't have been orange in the first place--it was supposed to be red!

Oh and yes--I uprezzed the image 300%, which is why it's as blurred as your M9 image which hasn't even got an AA filter!
 
Joseph S Wisniewski said:
But if the camera has a high metameric index (Foveon has a 0.199, according to their own data), even if you get a you can have a makeup that really looks good to a human eye or color film or a common Bayer camera, suddenly look really strange, too green or orange...
Click to expand...
According to "Eyeing the Camera: Into the Next Century, 2003, Richard F. Lyon and Paul M. Hubel Foveon, Inc.", the SD9 have a metameric index of 0.199. You're perfectly right about this. But is it a high one ?
According to the same paper, no. Quote from the publication :

"This early graphical analysis is what led us to conclude that the response curves that we could get from silicon color filtering would yield relatively small color errors, with suitable matrixing."

What allow them to said so ? Well, they provided comparaisons data. The calculate the metameric index of three other imaging devices (i'll let people run search if they want to know more about thoses). The results are the following (lower numbers show greater color accuracy) :

"Camera > Sensor > Metamerism index
Kodak DSC-460 > Kodak > 0.2974
Concord EyeQ > Agilent > 0.2873
Sigma SD9 > Foveon X3-F7 > 0.1999
HP 618 > Sony ICX-284 > 0.1802"
Notes : EyeQ Model undisclose, the HP is actually a scanner.

We shouldn't stop here because ; 1. The data is outdated (newer cameras might give totally different results) and 2. Maybe they just choose the worst cameras avaible for this comparaison (not far from the truth...).

The question we should ask ourself is ; general speaking, what is the metamerism index of digital cameras ? Is it far from the Sigma ones ?

According to DxOMark, ( http://www.dxomark.com/index.php/en/Learn-more/Understanding-DxOMark-Database/Measurements/Color-sensitivity ) :

"SMI [sensitivity metamerism index] is an index quantifying this property, and is represented by a number lower than 100 (negative values are possible). A value equal to 100 is perfect color accuracy, and is only attained when Luther-Ives conditions hold (which, as previously stated, never happens in practice). ... In practice, the SMI for DSLRs ranges between 75 and 85, and is not very discriminating. It is different for low-end cameras (such as camera phones) which typically have a SMI of about 40."

Now wait, 0.1999 versus 75 to 85 ? What's going on here ? ;)

Well, supposly, Lyon/Hubel used the same ISO standard (not that ISO ;)), yet the results measurements are obviously different. The calculation to equal thoses results is unknow to me, but fortunatly, Lyon/Hubel did it in another paper ("X3 Sensor Characteristics" - no date, online publication).
So what's the SMI of the sd9 (equalized to DxOMark measurement) ?
Drum-roll...

.... 91 ! Remember perfect score (100) just never happend, and the avreage score of DSRL is 80...

--
http://www.flickr.com/photos/maximebrousse/
 
Now that was what I was waiting for! I realized a couple of days ago that there has been a lot of talk about the theoretical limit of the different designs, but not about the real-world performance. So, the hypothetically the hard limit of 0.199 in the Foveon design doesn't matter much, if the actual accuracy is above the actual accuracy of Bayer sensors.

Now, of course, let's wait and see if someone can't poke a hole in your thesis.

EDIT: A couple of error.
Maxime B said:
Joseph S Wisniewski said:
But if the camera has a high metameric index (Foveon has a 0.199, according to their own data), even if you get a you can have a makeup that really looks good to a human eye or color film or a common Bayer camera, suddenly look really strange, too green or orange...
Click to expand...
According to "Eyeing the Camera: Into the Next Century, 2003, Richard F. Lyon and Paul M. Hubel Foveon, Inc.", the SD9 have a metameric index of 0.199. You're perfectly right about this. But is it a high one ?
According to the same paper, no. Quote from the publication :

"This early graphical analysis is what led us to conclude that the response curves that we could get from silicon color filtering would yield relatively small color errors, with suitable matrixing."

What allow them to said so ? Well, they provided comparaisons data. The calculate the metameric index of three other imaging devices (i'll let people run search if they want to know more about thoses). The results are the following (lower numbers show greater color accuracy) :

"Camera > Sensor > Metamerism index
Kodak DSC-460 > Kodak > 0.2974
Concord EyeQ > Agilent > 0.2873
Sigma SD9 > Foveon X3-F7 > 0.1999
HP 618 > Sony ICX-284 > 0.1802"
Notes : EyeQ Model undisclose, the HP is actually a scanner.

We shouldn't stop here because ; 1. The data is outdated (newer cameras might give totally different results) and 2. Maybe they just choose the worst cameras avaible for this comparaison (not far from the truth...).

The question we should ask ourself is ; general speaking, what is the metamerism index of digital cameras ? Is it far from the Sigma ones ?

According to DxOMark, ( http://www.dxomark.com/index.php/en/Learn-more/Understanding-DxOMark-Database/Measurements/Color-sensitivity ) :

"SMI [sensitivity metamerism index] is an index quantifying this property, and is represented by a number lower than 100 (negative values are possible). A value equal to 100 is perfect color accuracy, and is only attained when Luther-Ives conditions hold (which, as previously stated, never happens in practice). ... In practice, the SMI for DSLRs ranges between 75 and 85, and is not very discriminating. It is different for low-end cameras (such as camera phones) which typically have a SMI of about 40."

Now wait, 0.1999 versus 75 to 85 ? What's going on here ? ;)

Well, supposly, Lyon/Hubel used the same ISO standard (not that ISO ;)), yet the results measurements are obviously different. The calculation to equal thoses results is unknow to me, but fortunatly, Lyon/Hubel did it in another paper ("X3 Sensor Characteristics" - no date, online publication).
So what's the SMI of the sd9 (equalized to DxOMark measurement) ?
Drum-roll...

.... 91 ! Remember perfect score (100) just never happend, and the avreage score of DSRL is 80...

--
http://www.flickr.com/photos/maximebrousse/
Click to expand...
 
Joseph S Wisniewski said:
So, we build our machines (film and camera) like us, with three color vision that we try to make work like our own eyes. But our films and our cameras are imperfect "observers". This gives rise to something called a "failure of observer metamerism". (often shortened to simply "metamerism", although "metamerism" is "good", it is the "failure of metamerism" that is "bad").
Click to expand...
This is the reason, little understood, that aliens are little green men. Actually their skins have the same reflective properties as ours, but their eyes have a different spectral response. When their TV signals were intercepted, they were incorrectly decoded using our colour space.

--
Bob
 
1. The theorical limit still require references from my point of view (but let's give it time)

2. No thesis here, mere facts, with sources. What could happend; references being invalidated (been cautious about that). I should take this oportunity to correct the source of the second Lyon/Hubel reference ; it's actually been published (it's not just a online doc). http://sciencelinks.jp/j-east/article/200309/000020030903A0173461.php

4. This second paper actually explain why (sub-specification variation of ISO standard...) the first one results read different :

"Using IR and UV interference filters designed at the right cut-off wavelengths, we were able to achieve a metamerism index of 91 (using the draft standard technique - ISO 17321). Table 1 shows a comparison of several sensor technologies using the new metamerism technique. Unlike the previous technique (ISO draft standard ISO17321 WD4) as used in a previous publication (6), this index increases with more accurate color."
3. Sigma cameras making to DxOMark DB would make things easiers...
--
http://www.flickr.com/photos/maximebrousse/
 
CFynn said:
enos said:
Joseph S Wisniewski said:
The Nikon one has Foveon like color problems. The Fuji one cures everything: better color, no differential noise, but several of us (myself included) believe it can't be built economically.
Click to expand...
You've mentioned this Fuji sensor several times, would you mind describing how it works and what it does differently from Foveon? All I could find about it is one or two vague press releases.
Click to expand...
Probably the one described here:
http://www.faqs.org/patents/app/20090140123
Click to expand...
Panasonic has a similar patent. If either of them ever manages to market the resulting sensor, it will be interesting to see the IPR fight.
--
Bob
 
Great job !!
Maxime B said:
Joseph S Wisniewski said:
But if the camera has a high metameric index (Foveon has a 0.199, according to their own data), even if you get a you can have a makeup that really looks good to a human eye or color film or a common Bayer camera, suddenly look really strange, too green or orange...
Click to expand...
According to "Eyeing the Camera: Into the Next Century, 2003, Richard F. Lyon and Paul M. Hubel Foveon, Inc.", the SD9 have a metameric index of 0.199. You're perfectly right about this. But is it a high one ?
According to the same paper, no. Quote from the publication :

"This early graphical analysis is what led us to conclude that the response curves that we could get from silicon color filtering would yield relatively small color errors, with suitable matrixing."

What allow them to said so ? Well, they provided comparaisons data. The calculate the metameric index of three other imaging devices (i'll let people run search if they want to know more about thoses). The results are the following (lower numbers show greater color accuracy) :

"Camera > Sensor > Metamerism index
Kodak DSC-460 > Kodak > 0.2974
Concord EyeQ > Agilent > 0.2873
Sigma SD9 > Foveon X3-F7 > 0.1999
HP 618 > Sony ICX-284 > 0.1802"
Notes : EyeQ Model undisclose, the HP is actually a scanner.

We shouldn't stop here because ; 1. The data is outdated (newer cameras might give totally different results) and 2. Maybe they just choose the worst cameras avaible for this comparaison (not far from the truth...).

The question we should ask ourself is ; general speaking, what is the metamerism index of digital cameras ? Is it far from the Sigma ones ?

According to DxOMark, ( http://www.dxomark.com/index.php/en/Learn-more/Understanding-DxOMark-Database/Measurements/Color-sensitivity ) :

"SMI [sensitivity metamerism index] is an index quantifying this property, and is represented by a number lower than 100 (negative values are possible). A value equal to 100 is perfect color accuracy, and is only attained when Luther-Ives conditions hold (which, as previously stated, never happens in practice). ... In practice, the SMI for DSLRs ranges between 75 and 85, and is not very discriminating. It is different for low-end cameras (such as camera phones) which typically have a SMI of about 40."

Now wait, 0.1999 versus 75 to 85 ? What's going on here ? ;)

Well, supposly, Lyon/Hubel used the same ISO standard (not that ISO ;)), yet the results measurements are obviously different. The calculation to equal thoses results is unknow to me, but fortunatly, Lyon/Hubel did it in another paper ("X3 Sensor Characteristics" - no date, online publication).
So what's the SMI of the sd9 (equalized to DxOMark measurement) ?
Drum-roll...

.... 91 ! Remember perfect score (100) just never happend, and the avreage score of DSRL is 80...

--
http://www.flickr.com/photos/maximebrousse/
Click to expand...
--
Please visit my galleries at :
http://www.flickr.com/photos/yoicz/
or
http://www.naturephotos.dk/NaturePhotos_ejergalleri.php?menu=3&_Ejer=57

 
Maxime B said:
Joseph S Wisniewski said:
But if the camera has a high metameric index (Foveon has a 0.199, according to their own data), even if you get a you can have a makeup that really looks good to a human eye or color film or a common Bayer camera, suddenly look really strange, too green or orange...
Click to expand...
According to "Eyeing the Camera: Into the Next Century, 2003, Richard F. Lyon and Paul M. Hubel Foveon, Inc.", the SD9 have a metameric index of 0.199. You're perfectly right about this. But is it a high one ?
According to the same paper, no. Quote from the publication :

"This early graphical analysis is what led us to conclude that the response curves that we could get from silicon color filtering would yield relatively small color errors, with suitable matrixing."

What allow them to said so ? Well, they provided comparaisons data. The calculate the metameric index of three other imaging devices (i'll let people run search if they want to know more about thoses). The results are the following (lower numbers show greater color accuracy) :

"Camera > Sensor > Metamerism index
Kodak DSC-460 > Kodak > 0.2974
Concord EyeQ > Agilent > 0.2873
Sigma SD9 > Foveon X3-F7 > 0.1999
HP 618 > Sony ICX-284 > 0.1802"
Notes : EyeQ Model undisclose, the HP is actually a scanner.

We shouldn't stop here because ; 1. The data is outdated (newer cameras might give totally different results) and 2. Maybe they just choose the worst cameras avaible for this comparaison (not far from the truth...).

The question we should ask ourself is ; general speaking, what is the metamerism index of digital cameras ? Is it far from the Sigma ones ?

According to DxOMark, ( http://www.dxomark.com/index.php/en/Learn-more/Understanding-DxOMark-Database/Measurements/Color-sensitivity ) :

"SMI [sensitivity metamerism index] is an index quantifying this property, and is represented by a number lower than 100 (negative values are possible). A value equal to 100 is perfect color accuracy, and is only attained when Luther-Ives conditions hold (which, as previously stated, never happens in practice). ... In practice, the SMI for DSLRs ranges between 75 and 85, and is not very discriminating. It is different for low-end cameras (such as camera phones) which typically have a SMI of about 40."

Now wait, 0.1999 versus 75 to 85 ? What's going on here ? ;)

Well, supposly, Lyon/Hubel used the same ISO standard (not that ISO ;)), yet the results measurements are obviously different. The calculation to equal thoses results is unknow to me, but fortunatly, Lyon/Hubel did it in another paper ("X3 Sensor Characteristics" - no date, online publication).
So what's the SMI of the sd9 (equalized to DxOMark measurement) ?
Drum-roll...

.... 91 ! Remember perfect score (100) just never happend, and the avreage score of DSRL is 80...

--
http://www.flickr.com/photos/maximebrousse/
Click to expand...
thanks for this, very interesting indeed

but

paper and theory aside, sd9's color problem is not fabricated from thin air, most people who has seen its output will feel its color accuracy is worse than bayer's, this latest paper says otherwise actually gives me doubt about this paper and SMI's usefulness, reducing whole spectrum's color accuracy to a single number this SMI may not be so good telling the real world story.

if this paper measures SMI of sd9 as 75, and sd15 as 91, I will be much more encouraged :D
 
Maxime B said:
According to DxOMark, ( http://www.dxomark.com/index.php/en/Learn-more/Understanding-DxOMark-Database/Measurements/Color-sensitivity ) :

"SMI [sensitivity metamerism index] is an index quantifying this property, and is represented by a number lower than 100 (negative values are possible). A value equal to 100 is perfect color accuracy, and is only attained when Luther-Ives conditions hold (which, as previously stated, never happens in practice). ... In practice, the SMI for DSLRs ranges between 75 and 85, and is not very discriminating. It is different for low-end cameras (such as camera phones) which typically have a SMI of about 40."

Now wait, 0.1999 versus 75 to 85 ? What's going on here ? ;)

Well, supposly, Lyon/Hubel used the same ISO standard (not that ISO ;)), yet the results measurements are obviously different. The calculation to equal thoses results is unknow to me, but fortunatly, Lyon/Hubel did it in another paper ("X3 Sensor Characteristics" - no date, online publication).
So what's the SMI of the sd9 (equalized to DxOMark measurement) ?
Drum-roll...

.... 91 ! Remember perfect score (100) just never happend, and the avreage score of DSRL is 80...
Click to expand...
Well, let me quote the full list of scores from that online paper (which, Maxime omitted to mention, can be found here: http://billjanes1.home.comcast.net/~billjanes1/binary/KopieSensorChar.pdf ):

HP 715 85
MegaVision S2 88
Sigma SD9 91
HP 618 92

So let me get this:
1. According to DXOMark, the SMI for DSLRs ranges between 75 and 85.
2. The worst of the bunch in Hubel's paper tops the DSLRs' range of scores.

3. If Hubel was trying to make Foveon look good, he should have had plenty of DSLRs with

Instead, he chose a scanner line sensor (the HP 618, as you say), a video camera sensor and something else unknown, which we don't even have any reason to believe would have SMI as good as, let alone better than, still image sensors.

Clearly something is wrong ;) and since the same sensors were measured, with results in the same pecking order, it seems obvious to me that the results were simply recast in a different metric, rather than re-measured.

The Hubel paper said their measurements were according to the draft standard technique ISO 17321. No such qualification was given in the DXOMark page. If the discrepancy can't simply be put down to different errors in the measuring setups the two groups used, it could well be because the "draft ISO 17321 and the final ISO 17321 are different beasts.

Finally, as ns66 points out--if the SD9 had better SMI out of the sensor than all other DSLRs at the time--heck, even all non-Foveon DSLRs since--why couldn't it output accurate colours even with raw output, a desktop computer and all the computing power in-camera jpeg engines don't have? Keep in mind the comment on SMI at DXOMark:

"The underlying physics is that a sensor can distinguish exactly the same colors as the average human eye, if and only if the spectral responses of the sensor can be obtained by a linear combination of the eye cone responses. These conditions are called Luther-Ives conditions"

We know that Bayer cameras had been able to get away with linear combination for jpeg colour processing for quite a while with reasonably good results. If Foveon had better SMI than these cameras, it would have been able to do this even better than Bayer cameras--none of this faffing around with 24 or 124 patch colour space calibration for each scene would have been necessary to cudgel the Foveon into accurate colour response. In-camera jpeg output would have been simplicity itself!
 
Kendall Helmstetter Gelner said:
Joseph S Wisniewski said:
Kendall Helmstetter Gelner wrote:

If you're designing a camera, a 50% hike in processing, transfer, and storage requirements is a disaster.
Click to expand...
Thankfully that only makes camera design harder, not shooting it.
Click to expand...
No, it makes the camera either more expensive (50% faster processors, 50% larger and 50% faster memory cards) or slower. Making it slower affects how it's shot. Making it more expensive affects who shoots it, and for what purpose.
Kendall Helmstetter Gelner said:
You design a camera once and then build it for other people many times.
Click to expand...
Thank God you don't design stuff.
Kendall Helmstetter Gelner said:
Kendall Helmstetter Gelner said:
Kendall Helmstetter Gelner said:
Yes, if pointed at a subject of solid color with no edges.
Click to expand...
If pointed at any object that is locally monochrome, which is pretty much 100% of the real world.
Click to expand...
No object is "locally monochrome"
Click to expand...
Well, once you replace "no" with "most", sure.

So far, in two paragraphs, you've demonstrated flawed knowledge of business practices, design, and image processing.

I meant what I said, seriously, Kendall.

If you care about Sigma, at all, if you want them to do well, to thrive, you really need to just stop taking and go away. Every time you open your mouth, you make all the rest of the Sigma advocates, Sigma shooters, and, by association, Sigma itself, look like a bunch of drooling lunatics.

If you can't understand this, my pity.

--
Rahon Klavanian 1912-2008.

Armenian genocide survivor, amazing cook, scrabble master, and loving grandmother. You will be missed.

Ciao! Joseph

http://www.swissarmyfork.com
 
Joe0Bloggs said:
3. If Hubel was trying to make Foveon look good, he should have had plenty of DSLRs with
Click to expand...
Yes, all of the sensors used for comparison were small sensors that were effectively obsolete when the paper was published and certainly not representative of the real competition. The most obvious choice for a comparison that would have been easily available to them would the Canon D30 (3mp CMOS sensor). They also don't describe the demosaicing techniques they used.
Joe0Bloggs said:
Instead, he chose a scanner line sensor (the HP 618, as you say),
Click to expand...
No, the C618 used the ubiquitous 1/2" Sony CCD found in many in 2MP consumer cameras starting in 1998-1999.

--
Erik
 
Joe0Bloggs said:
Maxime B said:
According to DxOMark, ( http://www.dxomark.com/index.php/en/Learn-more/Understanding-DxOMark-Database/Measurements/Color-sensitivity ) :

"SMI [sensitivity metamerism index] is an index quantifying this property, and is represented by a number lower than 100 (negative values are possible). A value equal to 100 is perfect color accuracy, and is only attained when Luther-Ives conditions hold (which, as previously stated, never happens in practice). ... In practice, the SMI for DSLRs ranges between 75 and 85, and is not very discriminating. It is different for low-end cameras (such as camera phones) which typically have a SMI of about 40."

Now wait, 0.1999 versus 75 to 85 ? What's going on here ? ;)

Well, supposly, Lyon/Hubel used the same ISO standard (not that ISO ;)), yet the results measurements are obviously different. The calculation to equal thoses results is unknow to me, but fortunatly, Lyon/Hubel did it in another paper ("X3 Sensor Characteristics" - no date, online publication).
So what's the SMI of the sd9 (equalized to DxOMark measurement) ?
Drum-roll...

.... 91 ! Remember perfect score (100) just never happend, and the avreage score of DSRL is 80...
Click to expand...
Well, let me quote the full list of scores from that online paper (which, Maxime omitted to mention, can be found here: http://billjanes1.home.comcast.net/~billjanes1/binary/KopieSensorChar.pdf ):

HP 715 85
MegaVision S2 88
Sigma SD9 91
HP 618 92

So let me get this:
1. According to DXOMark, the SMI for DSLRs ranges between 75 and 85.
2. The worst of the bunch in Hubel's paper tops the DSLRs' range of scores.
Click to expand...
That is strange, I have to agree with you. But far from impossible ; remember, it's "just" color accuracy, not DR or noise.
Joe0Bloggs said:
3. If Hubel was trying to make Foveon look good, he should have had plenty of DSLRs with inferior 85 score to choose from. It would have been a more valid comparison too since Foveon sensors were first going into DSLRs
Click to expand...
He didn't write it with DP forums in mind. Be realistic about this.

[I had to edit your text for "inferior" or the formating would cut your sentence here]
Joe0Bloggs said:
Instead, he chose a scanner line sensor (the HP 618, as you say), a video camera sensor and something else unknown, which we don't even have any reason to believe would have SMI as good as, let alone better than, still image sensors.
Click to expand...
Well, this paragraph is kind of confus(ing), so need clarification :
HP 715 = 3 mp compact camera

MegaVision S2 = 4 megapixel 31 mm square CCD sensor with 15 µm pixel size - for digital back
SD9 = SD9 ;)
HP 618 = Scanner
Joe0Bloggs said:
Clearly something is wrong ;) and since the same sensors were measured, with results in the same pecking order, it seems obvious to me that the results were simply recast in a different metric, rather than re-measured.
Click to expand...
The sd9 and the HP 618 are common in both papers. The rest is not. Yes it was remeasured, according to them.
Joe0Bloggs said:
The Hubel paper said their measurements were according to the draft standard technique ISO 17321. No such qualification was given in the DXOMark page. If the discrepancy can't simply be put down to different errors in the measuring setups the two groups used, it could well be because the "draft ISO 17321 and the final ISO 17321 are different beasts.
Click to expand...
Could be. Just be aware that what they call "draft" is quite different from what draft mean in daily use. It basicly mean, "subject to change". Here's the HTML 5 draft for the fun of it : http://dev.w3.org/html5/spec/Overview.html

Between, ISO 17321 current statut is "90.92 International Standard to be revised"
Joe0Bloggs said:
Finally, as ns66 points out--if the SD9 had better SMI out of the sensor than all other DSLRs at the time--heck, even all non-Foveon DSLRs since--why couldn't it output accurate colours even with raw output, a desktop computer and all the computing power in-camera jpeg engines don't have? Keep in mind the comment on SMI at DXOMark:

"The underlying physics is that a sensor can distinguish exactly the same colors as the average human eye, if and only if the spectral responses of the sensor can be obtained by a linear combination of the eye cone responses. These conditions are called Luther-Ives conditions"

We know that Bayer cameras had been able to get away with linear combination for jpeg colour processing for quite a while with reasonably good results. If Foveon had better SMI than these cameras, it would have been able to do this even better than Bayer cameras--none of this faffing around with 24 or 124 patch colour space calibration for each scene would have been necessary to cudgel the Foveon into accurate colour response. In-camera jpeg output would have been simplicity itself!
Click to expand...
Well, here's the subjectivity/personnal experience again. I mean, you can have it both way. We either talk of validated measured results or of "I found the colors of whatever camera not to be accurate".

The problem is that it's your experience and your standard. It doesn't "mean" anything. Other have other standards ; DP1 hues are perfectly fine for image ressource for instance (IR being not the worst of the bunch as far as review site are concerns). For Gene Hack (Sigma forum), the SD9 is the only digital worth shooting with. I could go on and on (even using other brands).
That why we make mesureaments. So things are on the level. Objective.

And the validated one from the sd9 said 91. If you have validated published results that unvalidate it, I'm all hear.

--
http://www.flickr.com/photos/maximebrousse/
 
Joseph S Wisniewski said:
Kendall Helmstetter Gelner wrote:

No, it makes the camera either more expensive (50% faster processors, 50% larger and 50% faster memory cards) or slower. Making it slower affects how it's shot. Making it more expensive affects who shoots it, and for what purpose.
Click to expand...
Yes, somewhat more expensive, for the gain of better IQ. Thus has it always been.

It's not like processors to handle the data are the largest expense in a camera system. Honestly, you should know better - and in fact you do, but you cannot admit to anyone when you are wrong, as witnessed by whatever response you give to this.
Joseph S Wisniewski said:
Joseph S Wisniewski said:
You design a camera once and then build it for other people many times.
Click to expand...
Thank God you don't design stuff.
Click to expand...
Be glad to hear how that statement is wrong, since manufacturing is by definition the process of taking a single design and making multiple copies of it. Do you see where you start to lose the reader again by departing from how things really work?
Joseph S Wisniewski said:
Joseph S Wisniewski said:
No object is "locally monochrome"
Click to expand...
Well, once you replace "no" with "most", sure.
Click to expand...
If I have a blue object next to a red object, how can you avoid the possibility that one of the spatially seperated data points used to determine the final color of a thing will fall on the red side, and one on the blue? Thus, it is not "locally monochrome" simply by how the spatially separated sensor elements fall. This would seem to be obvious to anyone, and in fact it is - but you ignore it, again because you cannot admit you were wrong.
Joseph S Wisniewski said:
So far, in two paragraphs, you've demonstrated flawed knowledge of business practices, design, and image processing.
Click to expand...
Anyone reading knows that's false so all you've done is paint yourself as a bitter, angry person who can't admit his own failings.
Joseph S Wisniewski said:
I meant what I said, seriously, Kendall.
Click to expand...
So did I, you should heed my advice. You used to have a reputation as somewhat intelligent in regards to technical matters but with these series of responses (manufacturing is not mass production? Really?) you are seriously beclowning yourself. It's not too late to work to regain your reputation.
Joseph S Wisniewski said:
If you care about Sigma, at all, if you want them to do well, to thrive, you really need to just stop taking and go away. Every time you open your mouth, you make all the rest of the Sigma advocates, Sigma shooters, and, by association, Sigma itself, look like a bunch of drooling lunatics.
Click to expand...
Since I know that many people have been appreciative of my efforts over time (and not just Sigma users) this is just another thing you can't get right...
Joseph S Wisniewski said:
If you can't understand this, my pity.
Click to expand...
Oh right, so sad to be someone who has been correcting your mistakes for a decade so far, and decades to come. I don't need your pity since I'm leading a great life thanks. People only offer pity like that by way of an attack, how sad to stoop to such a lowbrow response to disagreement. Instead of really technical correction you offer only personal attacks and nonsense that leave even the non-technical people scratching heads over why you bothered posting.

Success is the best revenge, I always say. Instead of drowning in your pity I raise a glass to knowledge and understanding, my faithful companions for many years past, and many years to come.

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmadslr/user_home
 
Chato said:
I believe Joe is right in saying that an image from a Foveon chip is more likely to have poor color then one from a Bayer Pattern chip. Bayer is very consistant.
Click to expand...
Bayer is very consistent across a large area, where it falls down is around the edges where things come down to just a few pixels.

So overall the image looks great, BUT the image has a number of tiny flaws scattered throughout.

Now mostly they don't really matter much but I think they are in fact the reason why people talk about a "digital" look to images, because even though these small errors are not totally obvious, the subconscious notices things like detail falling off very suddenly and in a non-uniform way.
Chato said:
Where I do disagree with Joe, is that in my opinion the 90 percent of Foveon images that do not suffer from shift, are more accurate then the average Bayer pattern image. I shoot dawns twice a week, and the results of my DP2 are able to capture the weird colors better then my D2x.
Click to expand...
I would say exactly the same thing. That's why Joe's insistence that the Foveon situation is so much worse is puzzling, since it doesn't match with the reality of my shooting experience across multiple Foveon based cameras, and some Bayer ones.
Chato said:
And from looking at results over the years, Foveon based machines are making consistent progress in dealing with this problem. What the new SD1 will do is something we're just going to have to wait for.
Click to expand...
I totally agree, the Foveon progress seems slow, but it seems like this time they may have made a decent leap.

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmadslr/user_home
 
gaussian blur said:
Kendall Helmstetter Gelner said:
Joseph S Wisniewski said:
It's also why no one takes those color resolution charts that you've been talking about, for so many years, seriously.
Click to expand...
Since the color chart images reflect the effects of real world shooting...
Click to expand...
Nope. Colour charts do not reflect real world shooting conditions,
Click to expand...
Since I showed real world images from two cameras with the same effects, obviously they do. You're going to have to change your thinking when presented with proof, or else you'll just remain stuck I think.
gaussian blur said:
Put a B/W resolution chart and a colour resolution chart on the wall next each other and you will see a lot more detail on the B/W chart than on the colour one.
Click to expand...
Yes indeed, that's a real problem with bayer sensors. You'd not have that issue with a Foveon camera.
gaussian blur said:
Bonus points if you pick two colours with exactly the same luminance. That's why B/W charts are used to measure resolution and why eye doctors use B/W charts (other than colour blindness tests).
Click to expand...
Color blindness is an issue in distinguishing differences between some colors, but it hardly related to what happens if you shoot a blue on red chart. Which in the real world is the same as shooting fall leaves agains the sky. See how the real world comes into play when you start talking about color?
gaussian blur said:
gaussian blur said:
gaussian blur said:
You really don't want people to go hunting up images that break the Foveon, because there's a lot of examples out there.
Click to expand...
Well go find them man! And I can tell you why in fact whatever effect you see is preferred photographically over what would happen in the bayer case.
Click to expand...
Preferred by whom? You?
Click to expand...
In other words you can't even find any problem images. Thanks for the confirmation.
gaussian blur said:
gaussian blur said:
It is odd that all of these timeless Foveon dissenters simply cannot (or will not) actually produce real links showing issues when I can so readily do so with Bayer problems.
Click to expand...
Actually, they can and have.
Click to expand...
Which you and Joe can't seem to post. Again, thanks for the confirmation.
gaussian blur said:
More importantly, there is no such thing as a perfect image so no matter what image you look at, there is always going to be something wrong with it and that includes Foveon. It's guaranteed.
Click to expand...
Yes, that is the one thing you've got right so far.

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmadslr/user_home
 
Joe0Bloggs said:
Kendall Helmstetter Gelner said:
You may be tired of that one. How about the M9 then? Same issue.

Click to expand...
The railings are clearly defined in front of the runner and if there's a colour shift in the railings in front of the pink box, my eyes are hurting too much from the colour clash to see it! And what has lens CA got to do with the sensor?
Click to expand...
The railings are clearly defined in front of the runner - except for his legs! Note how they are ONLY clear in areas where the background is nearly black or white.

If it's CA at work, what lens gives you flesh colored CA?

This lack of consistency in image capture is the very thing I am talking about as an issue. In the M9 sample image at web size, I was wondering why the runner was on the other side of the railing.
Joe0Bloggs said:
Your images are also very blurred. Either they were out of focus, or you up-interpolated them.
Click to expand...
Yes, I should have mentioned it's a 200% upsizing (bicubic) just to make the issue clearer.

It is a matter of fact that if you go past the resolution limit of the sensor, colours will blur together, no matter what sensor technology you use.

An interesting supposition! If that were true, you'd see the same effect from white or black. Yet you do not:



In these images the railings are clearly defined.
Joe0Bloggs said:
This is from the DP1:

Click to expand...
First of all, thank you very much for bringing real images to the table in this discussion!

Indeed there is some color bleed through, but not an outright vanishing as with the railings against legs - instead the falloff is more natural.
Joe0Bloggs said:
See all the orange lines turning purple--even for the clearly defined semicircular lines? Oh what lousy colour resolution--from a Foveon sensor! Oh by the way, that shouldn't have been orange in the first place--it was supposed to be red!
Click to expand...
Well that's what happens when you shoot a DP-1 and overexpose reds, later cameras did not have this issue. Also the JPG conversion (in camera JPG I think used for the reviews) may have altered the tone a bit. But that image brings up an opportunity to present a 10MP shot from a Bayer camera ( the D60 ) of the same target (200 upsampling again):



Do you see how much more clearly the target was resolved in the camera that is supposedly 4.7MP?

Also in the Foveon image there was some bleeding, but note that it was smooth so that the semi-circles look like semi-circles from a distance. But in the 10MP bayer image, the red bands are suddenly interrupted at times by blue points.

That's why it's generally accepted that the real way to read out Foveon resolution numbers is to state something like:

14 MP Foveon (10MP bayer equivalent)
Joe0Bloggs said:
Oh and yes--I uprezzed the image 300%, which is why it's as blurred as your M9 image which hasn't even got an AA filter!
Click to expand...
Yes, again sorry for not mentioning it was upsized, though I generally stop at 200% so as not to make the transition too dramatic or introduce too many artifacts.

--
---> Kendall
http://InsideAperture.com
http://www.pbase.com/kgelner
http://www.pbase.com/sigmadslr/user_home
 
Kendall Helmstetter Gelner said:
Kendall Helmstetter Gelner said:
Kendall Helmstetter Gelner said:
Kendall Helmstetter Gelner said:
It's also why no one takes those color resolution charts that you've been talking about, for so many years, seriously.
Click to expand...
Since the color chart images reflect the effects of real world shooting...
Click to expand...
Nope. Colour charts do not reflect real world shooting conditions,
Click to expand...
Since I showed real world images from two cameras with the same effects, obviously they do. You're going to have to change your thinking when presented with proof, or else you'll just remain stuck I think.
Click to expand...
Human vision doesn't resolve chroma as well as it does luminance, which is why using a colour resolution chart is irrelevant and why Bayer works as well as it does.

You can demonstrate this by converting an image in Photoshop to Lab mode and blurring the ab channels. You won't notice a difference, even with a sizable radius (5-10 pixels). On the other hand, blur the L (luminance) channel even just a fraction of a pixel and it will be noticeable.
Kendall Helmstetter Gelner said:
Kendall Helmstetter Gelner said:
Put a B/W resolution chart and a colour resolution chart on the wall next each other and you will see a lot more detail on the B/W chart than on the colour one.
Click to expand...
Yes indeed, that's a real problem with bayer sensors. You'd not have that issue with a Foveon camera.
Click to expand...
Read what I wrote again. Here's a hint: I didn't say anything about Bayer or Foveon sensors.
Kendall Helmstetter Gelner said:
Kendall Helmstetter Gelner said:
Bonus points if you pick two colours with exactly the same luminance. That's why B/W charts are used to measure resolution and why eye doctors use B/W charts (other than colour blindness tests).
Click to expand...
Color blindness is an issue in distinguishing differences between some colors, but it hardly related to what happens if you shoot a blue on red chart. Which in the real world is the same as shooting fall leaves agains the sky. See how the real world comes into play when you start talking about color?
Click to expand...
Blue/red charts have absolutely nothing to do with shooting fall leaves or real world subjects. How absurd.
Kendall Helmstetter Gelner said:
Kendall Helmstetter Gelner said:
Kendall Helmstetter Gelner said:
Kendall Helmstetter Gelner said:
You really don't want people to go hunting up images that break the Foveon, because there's a lot of examples out there.
Click to expand...
Well go find them man! And I can tell you why in fact whatever effect you see is preferred photographically over what would happen in the bayer case.
Click to expand...
Preferred by whom? You?
Click to expand...
In other words you can't even find any problem images. Thanks for the confirmation.
Click to expand...
I can find countless problem images, however, I have far better things to do than play that game.
 
Kendall Helmstetter Gelner said:
Kendall Helmstetter Gelner said:
See all the orange lines turning purple--even for the clearly defined semicircular lines? Oh what lousy colour resolution--from a Foveon sensor! Oh by the way, that shouldn't have been orange in the first place--it was supposed to be red!
Click to expand...
Well that's what happens when you shoot a DP-1 and overexpose reds, later cameras did not have this issue. Also the JPG conversion (in camera JPG I think used for the reviews) may have altered the tone a bit.
Click to expand...
Excuses, excuses.

Foveon failed to properly reproduce the very thing you say it's good at, red/blue charts.
Kendall Helmstetter Gelner said:
But that image brings up an opportunity to present a 10MP shot from a Bayer camera ( the D60 ) of the same target (200 upsampling again):
Click to expand...
Big deal. So neither one can properly reproduce a red/blue chart. How often do you take photos of red/blue charts anyway? More often than you photograph that wall with varying red tones?

Bayer handles real world subjects very well, as millions of photos prove.
 
gaussian blur said:
Human vision doesn't resolve chroma as well as it does luminance, which is why using a colour resolution chart is irrelevant and why Bayer works as well as it does.

You can demonstrate this by converting an image in Photoshop to Lab mode and blurring the ab channels. You won't notice a difference, even with a sizable radius (5-10 pixels). On the other hand, blur the L (luminance) channel even just a fraction of a pixel and it will be noticeable.
Click to expand...
I think there is one flaw in your argument. Using a Gaussian blur on the croma or doing a (somewhat) ideal downscaling is perceptually similar, and corresponds to removing high-frequency detail.

The sampling grid of a Bayer sensor (or any image sensor, in fact) is far from an ideal Nyquist sampler. As long as you are not doing heavy pre-filtering, you will get creation of false detail (aliasing) for both Bayer and Foveon sensors in one form or another. Aliasing in the color channels can be annoying, just as noise in the color channels is annoying.

There are solutions to this (as evident in the massive amount of subjective "good" images available using Bayer sensors).
 
I think that practice has shown that we prefer images from Bayer sensors where the OLPF is tuned to suit the luminance resoslution, rather than the chrominance resolution. Meaning that we prefer having very sharp black&whites at the expense of having some aliasing in the colors.

Perhaps the perfect OLPF would filter more at red and green wavelengths, and let through more of the green spatial detail?
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top