Yellow on Quattro?

[Raist3d]

[No message]

 

[Raist3d]

Anyway, just thinking out loud. Would be interesting to see what happens with colors like red on purple, etc.

A link to my color resolution comparison (Just in case you missed that thread)

color resolution

Thanks. That also shows the resolution differences.

 

[PrebenR]

I tried a few things but effectively yes, the issue does show up. This is SPP 6.0.3. This is the current version of Sigma SPP available at the Sigma website for download (at least from USA links), as of July 24, 2014.

Huh? Strange, why don't they have 6.0.4? Lazy to update?

Here is a screen grab. Notice I put the noise reduction all the way down. If there is a setting I am missing please point it out. Thanks!

PrebenR- I think something interesting to noice- notice the kind of blue that is against the yellow. Seems like a cyan blue, closer to the top layer response than the blue you had. Maybe that's what makes it more problematic.

It looks very much like the example I had from that label. Would be interesting to experiemnt with.

I still think there is some NR/interpretation when the SPP processes the file, but hard to say.

 

[PrebenR]

The Merrill won't have this (as you see it). Preben was saying they are using an older version of SPP, certainly newer SPP could deal with this better. But as I said many times, it is a FACT, not a guess, that the Quattro has less chroma resolution- even an exFoveon engineer confirmed this.

The problem though is that you would get this effect at the border only.

Not necessarily- you can get this at any point of color transitions that are abrupt within for example cloth threads.

But that is a border as I say.

Other tones can be faked/interpolated and they would look close but not like an X3 capture.

If it is purly chroma and not some processing issue. Then you are talking about 2 pixels. So for this to happen in the banners (both of them show this), then in my mind it is down to processing.

Not sure if you are talking about now of your shot or the "beer garden" shot.

Not my shots.

I think this would bother me more on the Quattro if there wasn't any true X3 capture mode, but there is. I would be happy shooting at 4.9MP and using the higher resolution for B&W.

But do you then say that any 20Mp Bayer camera is in effect a 5Mp camera? As it doesn't have more blue or red photosites?

No. That's not what I said. What I said is if I want a true X3 capture with full color acuity on the Quattro, I would just shoot at 4.9MP.

So you want the 4 measurements of the top layer to be combined? How does it make it more acute? Not sure I follow, but I can test it.

According to the docs it's one big. Keep in mind, I am not talking about Bayer, I am talking about Quattro vs previous Foveon. Basically if one unique characteristic of Foveon is not quite there and it looks more towards an AAless Bayer in some ways, some uniqueness of it is lost.

Whatever, I'm fed up discussing the Bayer notion that is stuck in peoples head.

 

[Scottelly]

I think we need to see a plurality of shots but the other link sure shows the color resolution loss (the wheels).

--
Raist3d/Ricardo (Photographer, software dev.)- I photograph black cats in coal mines at night...
“The further a society drifts from truth the more it will hate those who speak it.” - George Orwell

Here is an image that I am posting full-size. It is a bunch of yellow characters on a blue background in 8.5x11 size at 300 dpi. This should be sufficient for someone to print out on glossy paper and photograph with both cameras from a distance of about 5 or 10 feet (maybe both). If we see a bunch of blur in the Quattro image vs. the Merrill image, and there is no noise reduction being added into the mix, then we have reproduced the problem, and we will need to figure out what is going on. If not, then we will know that those problems were aberrations and not the norm . . . right?




For some reason the image has artifacts in the preview you see above. View it at 100%.

Please, anyone with both a DP2 Quattro and a DP2 Merrill, print this on a full-size sheet of paper and photograph it with both cameras in raw mode. Please post the resulting photos, processed without any noise reduction added (multiple versions would be interesting to see).

 

[Scottelly]

Here are better, smaller (in the number of kilobytes) GIF versions of the image above and the same image with a lighter blue background:




Same image as above, but in GIF file format - much cleaner and less than quarter the data size.




Here is the same image with a lighter blue background - in GIF image format.

Someone please print these and shoot them with your Quattro!

 

[Scottelly]

In "Flat View" you can see two test images I posted, just above this post. Someone with a DP2 Quattro please shoot a photo of each of those test images, and post it (no noise reduction please!), so we can see if this phenomenon always happens with yellow on blue.

 

[Raist3d]

I tried a few things but effectively yes, the issue does show up. This is SPP 6.0.3. This is the current version of Sigma SPP available at the Sigma website for download (at least from USA links), as of July 24, 2014.

Huh? Strange, why don't they have 6.0.4? Lazy to update?

Here is a screen grab. Notice I put the noise reduction all the way down. If there is a setting I am missing please point it out. Thanks!

PrebenR- I think something interesting to noice- notice the kind of blue that is against the yellow. Seems like a cyan blue, closer to the top layer response than the blue you had. Maybe that's what makes it more problematic.

It looks very much like the example I had from that label. Would be interesting to experiemnt with.

The blue looks different time. Also to experiment: cyan blue and low intensity. That may confuse the algorithm more.

I still think there is some NR/interpretation when the SPP processes the file, but hard to say.

Why? There rest of the areas look pretty sharp

 

[PrebenR]

The blue looks different time. Also to experiment: cyan blue and low intensity. That may confuse the algorithm more.

Will do.

I still think there is some NR/interpretation when the SPP processes the file, but hard to say.

Why? There rest of the areas look pretty sharp

No, I meant the algorithm that is used to develop the Quattro RAWs. Not what one sets in SPP.

I have done some testing and I have found that one can recreate this issue, but at the moment it is only for some certain colours. If I change the background a bit then the yellow is fine again.

I'll post back in the forum when I have collected more data. To me it looks like the algorithm gets something wrong in some particular circumstances.

 

[PrebenR]

In "Flat View" you can see two test images I posted, just above this post. Someone with a DP2 Quattro please shoot a photo of each of those test images, and post it (no noise reduction please!), so we can see if this phenomenon always happens with yellow on blue.

No it doesn't always happen. I have managed to get the issue, but I have not tried your samples. I'll do much more experiments on this and then I'll post my findings.

 

[Raist3d]

Do keep in mind "the algorithm getting it wrong in specific situations" is exactly what I would expect. The reason changing color / luminance intensity affects is because its going to play with the thresholds of whatever they are doing to get a red or green instead of a blue out of the top layer. You can't interpret the top layer as two colors at once a the same photosite-spatial location. All you have to go by is the three voltages from the three layers with two at 1/4th the resolution of the top as they respond to wavelengths in light spectrum

For the top layer given the spectral response curve a weak intensity cyan may respond with almost the same voltage as a somewhat strong magenta-pink. That's part of what makes this tricky.

 

[PrebenR]

OK if you know the answer, then I don't need to test this further? The problem goes beyond the 2x2.

 

[Raist3d]

OK if you know the answer, then I don't need to test this further?

That's not what I said. I mentioned what can be tricky situations or I think should be. I did not say you should stop testing (but stop if you want, I mean, by all means).

The problem goes beyond the 2x2.

 

[Raist3d]

Yes, we are on the same page.

 

[PrebenR]

OK if you know the answer, then I don't need to test this further?

That's not what I said. I mentioned what can be tricky situations or I think should be. I did not say you should stop testing (but stop if you want, I mean, by all means).

Relax, I'm going to do a lot of tests. But I'll wait until I finish to discuss it further.

 

[PrebenR]

Sounds like the issue (as the other post shows) may perhaps have to do with transitions of blue and colors requiring from the layers below. It's interesting.

--

Out lightwriting with Sigma
«Kaos»

 

[Raist3d]

OK if you know the answer, then I don't need to test this further?

That's not what I said. I mentioned what can be tricky situations or I think should be. I did not say you should stop testing (but stop if you want, I mean, by all means).

Relax, I'm going to do a lot of tests. But I'll wait until I finish to discuss it further.

 

[Raist3d]

Thank you for discussing out musings, and the subjects user discussion in a civil manner. I appreciate it.
--
Raist3d/Ricardo (Photographer, software dev.)- I photograph black cats in coal mines at night...
“The further a society drifts from truth the more it will hate those who speak it.” - George Orwell

 

[Raist3d]

Very interesting. One shot nominal, the other same frame and all but with a yellow filter.

http://ichigo-up.com/Sn4/download/1405776566.X3F

http://ichigo-up.com/Sn4/download/1405772202.X3F

Looks like those "beer gardeners" have been busy Converted by me below, SPP 6.0.4

It's interesting the resolution for the Yellow sign comes back with a Yellow filter on top of the lens, but now there's a bit of color moire on the top fence of the building. Haven't looked at all the pic carefully to check for other artifacts.

First Foveon with color moire apparently

A few predicted this could happen, but the color moire will sure be of a different characteristic than classic AAless Bayer- and it is. Alternates with Yellow and Luma.

I don't think this would be an issue for a landscape (unless it has a fence and you are using a color filter ). Fashion with weird textiles? Good question. Probably pretty rare too.

I can only think the original shot with that particular blue at that particular intensity makes it hard to discern high resolution red/green-> yellow.

--
Raist3d/Ricardo (Photographer, software dev.)- I photograph black cats in coal mines at night...
“The further a society drifts from truth the more it will hate those who speak it.” - George Orwell

 

[bford]

Very interesting. One shot nominal, the other same frame and all but with a yellow filter.

http://ichigo-up.com/Sn4/download/1405776566.X3F

http://ichigo-up.com/Sn4/download/1405772202.X3F

Looks like those "beer gardeners" have been busy Converted by me below, SPP 6.0.4

It's interesting the resolution for the Yellow sign comes back with a Yellow filter on top of the lens, but now there's a bit of color moire on the top fence of the building. Haven't looked at all the pic carefully to check for other artifacts.

First Foveon with color moire apparently

A few predicted this could happen, but the color moire will sure be of a different characteristic than classic AAless Bayer- and it is. Alternates with Yellow and Luma.

I don't think this would be an issue for a landscape (unless it has a fence and you are using a color filter ). Fashion with weird textiles? Good question. Probably pretty rare too.

I can only think the original shot with that particular blue at that particular intensity makes it hard to discern high resolution red/green-> yellow.

--
Raist3d/Ricardo (Photographer, software dev.)- I photograph black cats in coal mines at night...
“The further a society drifts from truth the more it will hate those who speak it.” - George Orwell

interesting. thanks for posting.