Pages: 1 [2]   Go Down

Author Topic: Getting the A7R3 files closer to IQ100 back outputs?  (Read 4096 times)

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11088
    • Echophoto
Re: Getting the A7R3 files closer to IQ100 back outputs?
« Reply #20 on: January 21, 2018, 09:17:56 PM »

Hi,

I guess that "Narikin" perhaps wants to be able to mix images shot with the different cameras.

You are probably right. I did a lot of experimenting on the samples I downloaded from Imaging Resource.

Best regards
Erik


...
Bottom line: the Sony is a VERY good small-format camera (and very good from a performance per dollar value perspective). It's small and light and has good video features; it's a tremendous little camera. But I'm honestly a little confused why you'd expect it to match the image-quality performance of your 16 bit full-frame 645 Phase One camera.
Logged
Erik Kaffehr
 

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11088
    • Echophoto
Re: Getting the A7R3 files closer to IQ100 back outputs?
« Reply #21 on: January 21, 2018, 09:20:29 PM »

Hi Wayne,

Sloppy use of language. Thanks for making the point!

Best regards
Erik

I assume you mean for a Phase One back, and not for Phase One? Just sounds a little odd, perhaps might confuse some?
Logged
Erik Kaffehr
 

narikin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1121
Re: Getting the A7R3 files closer to IQ100 back outputs?
« Reply #22 on: January 21, 2018, 09:22:13 PM »

Thanks Erik,

Well, exactly.

Open this layered TIFF in ps, select the base layer IQ3100, then toggle on off layer #3 - the A7R3 Generic - and watch the yarn rainbow weaken dramatically.

Why so? It's beyond my knowledge to offer an answer. The sensor is same generation roughly, smaller of course, same C1 processing, so why does the color drop away?
Logged

Doug Peterson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3805
    • http://www.doug-peterson.com
Re: Getting the A7R3 files closer to IQ100 back outputs?
« Reply #23 on: January 21, 2018, 09:23:38 PM »

Thanks for the input Doug. And for Puddles Pity Party, which was great to listen to, but nope, not better!

Have to point out it's moot who's the original and who's the cover version when it comes to sensors and processing. Sony make the sensors, as you very well know. Not Phase. Yes, Phase do great things with them, but Sony is the original manufacturer, and have the engineers & scientists in house. So...

The analogy is imperfect (as most analogies are).

A [Camera] is much more than a [Sensor]. Sony through C1 or LR is not as "native" as P1 through C1.


And to clarify: I'm not expecting it to be the same, I just was expecting a smaller resolution version of the MF files, but: nope. That's what's confusing (and saddening) me. These are not competitive products, they complement each other quite perfectly otherwise, I'd love for the output files to be broadly comparable other than scale, but I just can't get there. Sigh.

It's a very reasonable assumption to make that small-format Sony is just like medium-format Phase, just smaller. It's concise. It's logical to some extent.

Unfortunately it's just not true. As you're finding, Phase One has both hardware/software/profiling under one roof and prioritizes image quality over all else, and it shows in the end result.

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11088
    • Echophoto
Re: Getting the A7R3 files closer to IQ100 back outputs?
« Reply #24 on: January 21, 2018, 10:00:36 PM »

Hi,

If you enable layer two, that is A7rIII_EKr and toggle layer 3 that is A7rIII_C1_Generic you can see the difference caused by profiles.

You can repeat the experiment using layer 1 (IQ3100 with C1 flash profile) and layer 4 (the Lumariver profile I generatted for the IQ3100MP).

So profiles matter...

We can compare colour rendition between Sony A7rIII and IQ3100MP if we activate layer 2 and layer 4 and toggle layer 4. Those are both using my profile. So that should be an estimate for the colour differences in sensor CFAs.

Best regards
Erik

Thanks Erik,

Well, exactly.

Open this layered TIFF in ps, select the base layer IQ3100, then toggle on off layer #3 - the A7R3 Generic - and watch the yarn rainbow weaken dramatically.

Why so? It's beyond my knowledge to offer an answer. The sensor is same generation roughly, smaller of course, same C1 processing, so why does the color drop away?
Logged
Erik Kaffehr
 

algrove

  • Full Member
  • ***
  • Offline Offline
  • Posts: 240
Re: Getting the A7R3 files closer to IQ100 back outputs?
« Reply #25 on: January 22, 2018, 01:40:03 AM »

I know others besides myself who are not pleased with Sony A7R3 results when comparing IQ3100 files versus a7R3 files. Double the resolution and another $40k must do something to make them better files. Let's hope so anyway.
Logged

Jack Hogan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 701
    • Hikes -more than strolls- with my dog
Re: Getting the A7R3 files closer to IQ100 back outputs?
« Reply #26 on: January 22, 2018, 06:05:18 AM »

Thanks Erik,

Well, exactly.

Open this layered TIFF in ps, select the base layer IQ3100, then toggle on off layer #3 - the A7R3 Generic - and watch the yarn rainbow weaken dramatically.

Why so? It's beyond my knowledge to offer an answer. The sensor is same generation roughly, smaller of course, same C1 processing, so why does the color drop away?

The reason is quite simple: a good many of the tones in the yarn (and the fabric elsewhere) are out of gamut in both cameras in Adobe RGB - so it's a crapshoot as to what the original colors looked like.  Other than that both cameras show a comparable color accuracy score under this artificial D50 light (SMI of about 86).

Jack
Logged

Jack Hogan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 701
    • Hikes -more than strolls- with my dog
Out of Gamut
« Reply #27 on: January 22, 2018, 08:51:10 AM »

And here are the Adobe RGB out-of-gamut images: the colors refer to the channel that is blocked (R,G, B - white means highlight clipping).  There are 473K tones blocked in the IQ3-100 vs 173K in the a7rIII, I would call it even given the different file sizes.  The latter appears to be slightly better in the yellow hues while the former in the reds.  As mentioned earlier, as far as in-gamut tones are concerned the two cameras appear to be about equivalent.



So viewing these D50 captures one really seems to learn more about how the relative profiles choose to paint-in out-of-gamut tones rather than the true capabilities of the hardware.  Operator error is always a possibility where I am concerned but I wonder if anyone has ever raised this issue with IR.

Jack
« Last Edit: January 22, 2018, 09:17:31 AM by Jack Hogan »
Logged

Doug Peterson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3805
    • http://www.doug-peterson.com
Re: Out of Gamut
« Reply #28 on: January 22, 2018, 09:21:20 AM »

And here are the Adobe RGB out-of-gamut images

Why would you measure the in-gamut and out-of-gamut in the images converted to Adobe RGB. Would it not make more sense to look at the files in their respective original color space (C1 allows you to embed the original profile when outputting to TIFF or JPG to facilitate such investigations).

Jack Hogan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 701
    • Hikes -more than strolls- with my dog
Re: Out of Gamut
« Reply #29 on: January 22, 2018, 10:28:51 AM »

Why would you measure the in-gamut and out-of-gamut in the images converted to Adobe RGB. Would it not make more sense to look at the files in their respective original color space (C1 allows you to embed the original profile when outputting to TIFF or JPG to facilitate such investigations).

Hello Doug,

Because I thought 99% of the folks here would be trying to judge color in these images by looking at them on their monitors, 99% of which only cover Adobe RGB or less.  So here is a question: why would a commercial site have a target with so much of it outside of Adobe RGB?  I can't imagine what the vast majority of people - with sRGB coverage at best - think they are judging as far as color is concerned ;-)

And I am not sure what the original color space of the files is.  My images are rendered 100% linearly with no adjustments/profiles whatsoever by multiplying the white balanced raw data by the optimized compromise color matrix for the individual capture as setup.  Then counting blocked (less than zero, out-of-gamut) and clipped (specular highlights in this case) pixels in the chosen destination color space.  The beauty of the linear color approach is that it speaks quantitatively to the capabilities of the hardware - before profiles and qualitative processing enter the picture to make up for its weaknesses and/or produce more pleasing results.  Which doesn't mean that as a landscape photographer I do not appreciate a well researched, robust profile to render pleasing images with a minimum of fuss.  But when trying to measure the capabilities of the hardware, profiles take a back seat.

Jack
Logged

opgr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2079
Re: Out of Gamut
« Reply #30 on: January 22, 2018, 10:48:31 AM »

My images are rendered 100% linearly with no adjustments/profiles whatsoever by multiplying the white balanced raw data by the optimized compromise color matrix for the individual capture as setup. 

But how do you set the multipliers? If you take the white patch and map it linear to 90% then the consequence is that a lot of other colors might get clipped by simple overexposure due to method, and not by colorconversion. It seems rather odd that the yarn would be out-of-adobegamut...
Logged
Regards,
~ O ~

Jack Hogan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 701
    • Hikes -more than strolls- with my dog
Re: Out of Gamut
« Reply #31 on: January 22, 2018, 11:10:21 AM »

But how do you set the multipliers? If you take the white patch and map it linear to 90% then the consequence is that a lot of other colors might get clipped by simple overexposure due to method, and not by colorconversion. It seems rather odd that the yarn would be out-of-adobegamut...

Hi Oscar,

I know, I was surprised myself about the yarn. I went through all this because I was flipping back and forth on the images of the yellow thread in PS (my render) and I noticed that blue was showing  a lot of zero values.  Then I looked around and I saw a lot more blocked tones.  The process I follow is outlined in this article.  The routine spits out the optimized matrix for the white balanced raw data as-is and it can be used as-is.  It assumes that the third CC neutral patch from the right has its published value of about L*=50, a*=0, b*=0.  In this case only specular reflections are clipped.  If brightness is reduced by a stop before applying the matrix the out of gamut values remain virtually unchanged.

Jack

BTW the lighting is different in character in the two images, you can see that in the shadows.  One is more diffuse than the other.
Logged

Doug Peterson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3805
    • http://www.doug-peterson.com
Re: Getting the A7R3 files closer to IQ100 back outputs?
« Reply #32 on: January 22, 2018, 11:24:28 AM »

What colors are in/out of gamut in the original raw file (in the original profile) is what really matters.

A savvy user can then decide whether and in what way to ensure the final output file is within gamut for a given use in a given profile (sRGB, Adobe 1998, ProPhoto, a particular printer etc).

Assuming of course your worried about the practical rather than the academic.

Jack Hogan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 701
    • Hikes -more than strolls- with my dog
Re: Getting the A7R3 files closer to IQ100 back outputs?
« Reply #33 on: January 22, 2018, 12:10:21 PM »

What colors are in/out of gamut in the original raw file (in the original profile) is what really matters.

I am not familiar with C1 or its terminology Doug.  The 'original' profile I assume is the Profile Connection Space, i.e. XYZ, so if the colors are known to be out of Adobe RGB gamut while there and one wishes to finally display the image in such a color space then someone, manually or automatically, needs to physically take the out-of-gamut tones and bring them in-gamut - because it's better to smooth out the blocking, right?  And the instant one starts twisting tones around by applying a profile one loses the ability to say that one camera's color is better than another by looking at the rendered image: too many changes applied, designed to give a certain 'look', sometimes far removed from the tones actually produced by the sensor.  Of course one can say anything anecdotally, as in 'I prefer the way C1 renders my raw files than LR'.

On the other hand to be able to say that one camera produces better color than another one needs to roll up one's sleeves and look under the hood where the engine can be seen running before the profile is applied.  But I know you know all this, having watched some of your webinars :)

Jack
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11088
    • Echophoto
Re: Getting the A7R3 files closer to IQ100 back outputs?
« Reply #34 on: January 22, 2018, 12:24:04 PM »

Hi Jack,

Thanks for that info. I have tested generating som ICC profiles with LumaRiver profile designer, and my files are quiet similar using the LumaRiver profiles.

I have also noticed that the test target is similar but not identical.

So, what I see is IQ3100 is pretty similar to A7rII with profiles generated by Lumariver but quiet different with the "Generic" profile built into C1. Sort of indicates that hardware capabilities are similar, except the IQ3100MP sensor being larger and having more pixels but the profiles by Phase One are different.

If I use a colour profile generated by LumaRiver with Capture One or LR, I get similar colours except on the blue yarn.

Best regards
Erik

The reason is quite simple: a good many of the tones in the yarn (and the fabric elsewhere) are out of gamut in both cameras in Adobe RGB - so it's a crapshoot as to what the original colors looked like.  Other than that both cameras show a comparable color accuracy score under this artificial D50 light (SMI of about 86).

Jack
Logged
Erik Kaffehr
 

DougDolde

  • Full Member
  • ***
  • Offline Offline
  • Posts: 150
    • Images of the American West
Re: Getting the A7R3 files closer to IQ100 back outputs?
« Reply #35 on: January 22, 2018, 12:46:45 PM »

I don't know about Sony or the 100 megapixel Phase One backs but I sold off my IQ180, body and lenses and replaced them with a Nikon D850 and three Zeiss ZF.2 lenses, 21mm, 28mm and 85mm.

I am blown away with the quality. Sure it's less resolution but it's plenty.  The DR, low light capability, Live View focus peaking, and it's much lighter weight and field friendly. 

Plus I can still process the files in Capture One Pro.  The total outlay was just under $6000

Take that Phase One
Logged

Jack Hogan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 701
    • Hikes -more than strolls- with my dog
Re: Getting the A7R3 files closer to IQ100 back outputs?
« Reply #36 on: January 22, 2018, 01:18:04 PM »

So, what I see is IQ3100 is pretty similar to A7rII with profiles generated by Lumariver but quiet different with the "Generic" profile built into C1. Sort of indicates that hardware capabilities are similar, except the IQ3100MP sensor being larger and having more pixels but the profiles by Phase One are different.

Ok, makes sense, thanks Erik.  BTW I checked DPR's D65 studio scene with a a7Riii capture and it's a different story.  Just a few bits here and there are out of aRGB gamut (some paint tubes etc.) but most of it is in-gamut.
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11088
    • Echophoto
Re: Getting the A7R3 files closer to IQ100 back outputs?
« Reply #37 on: January 22, 2018, 05:10:25 PM »

Hi Jack,

These images have been converted to sRGB, in order to simplify viewing in browsers:

First, IQ3100 processed in C1 with Flash profile:


Next is same image, but processed in C1 with a LumaRiver profile:


Up next is A7rIII processed in C1 with a Lumariver profile:


The last one is A7rIII processed with C1-s built in profile (ArrIII generic):


The differences are much more obvious if you stack the images in Photoshop:

http://echophoto.dnsalias.net/ekr/Articles/C1Stuff/LayeredComparison.tif

Best regards
Erik

Ok, makes sense, thanks Erik.  BTW I checked DPR's D65 studio scene with a a7Riii capture and it's a different story.  Just a few bits here and there are out of aRGB gamut (some paint tubes etc.) but most of it is in-gamut.
Logged
Erik Kaffehr
 
Pages: 1 [2]   Go Up