Capture One on the A7RII giving me disappointing results

[tho_mas]

It's not clear to me in how far this capture is complicated to handle.
Just select the "extra shadow" film curve in base characteristics, push exposure by 0.6EV and Shadow in the HDR tool to 45 and you'll have all the image content within the histogram (except for some specular highlights). From there on adjust the image by your liking.
Best practise to exchange settings is to use the original raw file (the arw file) and pack the image as "EIP". This way we can open the RAW with your C1-settings and can send you back the RAW files with our modified settings...

[Robert Ardill]

It does appear that C1 isn't too happy with dng files - the as-shot white balance between the dng and arw files is wildly different.  LR shows the same white balance with the dng and arw files (4750, +5), whereas C1 shows 5628,-1.6 for the arw and 4978, 16.9 for the dng. Why there are such differences between the two converters with the as-shot white balance is a bit strange.

I've re-developed the file using the arw original in C1 with the tone curve set to Linear as suggested by Bart.  The result is certainly much better  with pretty much equivalent detail in the C1 and LR versions.  The look of the images is certainly different, with LR giving a more HDR-look ... a question of taste, but I prefer the C1 version as I've developed it as it looks more natural. However it would be easy to achieve a similar look in LR ... and no doubt equally easy to get a more HDR-look in C1.

I was wrong in thinking that C1 applies luminance noise reduction even with the luminance slider set to zero ... at least there is no evidence of that using the Linear tone curve with the arw original.

I only used basic tonal adjustments. Except for color noise reduction all other settings were zeroed.

All of the files are available here (please re-download the dng file if you have already downloaded it as I've saved the development settings to it in this version):

 DNG File with LR adjustments   

 EIP File with C1 adjustments   

 ARW original   

 Rendered TIF with C1 adjustments   

 Rendered TIF with LR adjustments     

Cheers,

Robert           

[Robert Ardill]

P.S. I'll post my conversion when questions about profile and creative intent are more clear to me.

Hi Bart - I'm happy to go with the linear tone curve (if that is what you mean).  Regarding creative intent, in this case I have none. I was just looking at detail, artifacts, local contrast, shadow and highlight recovery. As before, with luminance noise reduction and sharpening off.

Cheers

Robert

[Bart_van_der_Wolf]

Hi Bart - I'm happy to go with the linear tone curve (if that is what you mean).

Yes, especially if you want to avoid blowing highlights (the film curve suggests to do that even if the file was ETTR). A linear tonecurve also retains more highlight detail, such as in clouds.

Regarding creative intent, in this case I have none. I was just looking at detail, artifacts, local contrast, shadow and highlight recovery. As before, with luminance noise reduction and sharpening off.

What I was l aiming to get was something close to what you were pleased with in the LR rendering, not what I would have made of the image without knowing the original scene and mood. Now that the Original Camera Raw is available I'll use that.

Cheers,
Bart

[Robert Ardill]

What I was l aiming to get was something close to what you were pleased with in the LR rendering, not what I would have made of the image without knowing the original scene and mood. Now that the Original Camera Raw is available I'll use that.

Cheers,
Bart

Hi Bart,

Here is the file developed in LR to be close to what I recall the scene being.  I've lightened the blue and desaturated the red (to reduce the HDR look) and added a touch of clarity.  I've added no sharpening (I would do a 1/150 sharpening in FM probably).

TIF -  LR-Developed

Cheers,

Robert

[AlterEgo]

Yes, especially if you want to avoid blowing highlights (the film curve suggests to do that even if the file was ETTR). A linear tonecurve also retains more highlight detail, such as in clouds.

as I illustrated in my C1 conversion above ( left part with OP's settings embedded in DNG he posted originally, right side C1 conversion with film curve set to "auto", which selects "film standard" curve - unless your profile for C1 has a specific instruction to select recorded in 'P1FC' private tag in icc/icm camera profile - see "repro" profiles for C1 CH edition for example ) there is absolutely no need to use linear curve to retain more highlight details for that particular shot - there are absolutely the same level of details as in his original ACR/LR rendering (that was recorded in DNG)... you just use contrast slider to bring it back closer to linear curve... and I used a camera profile compensated for C1 'transfer function' for A7R2 camera.

[Robert Ardill]

I've checked out C1's processing of dng/arw and there are differences. 

The most obvious is the wrong white balance value (dng: 5131/16.7 arw: 5812,-1.4) although the actual white balance is identical (comparing the TIFFs). So this seems just to be a display issue.

Much more concerning is that there is a small but noticeable difference in the rendered TIFFs.  In order to see this I placed the files as layers in Photoshop with the top layer blend mode set to Difference. With a levels adjustment applied it's clear that the files are different. However the differences are slight.

LR renders the arw and dng identically and the white balance reading on both is 4850,-6.

(RT shows different white balance readings also: arw: 5276, 1.085 dng: 4924, 1.041)

I'll report the issue to Phase One.

Cheers,

Robert

[tho_mas]

Robert ... just a short note on your settings in C1: color noise reduction is set to 50 ... way too high for the A7R2 even at ISO400 and further pushed in C1.

Besides I agree with "AlterEgo" regarding the linear response curve.... it's not needed for this capture to preserve highlight details.
Generally I also do use the linear response curve ... but not as my standard... only for certain captures when need is (I rarely shoot subjects with super high dynamic range, though). When you use the linear curve in base characterics you should be familar with using the curves tool. When you use the linear curve and only work with the contrast slider and the levels tool you may end up with "flat" and "dull" looking areas somewhere in your photos... sometimes things may look artificial (not film-like).

[Robert Ardill]

Robert ... just a short note on your settings in C1: color noise reduction is set to 50 ... way too high for the A7R2 even at ISO400 and further pushed in C1.

Besides I agree with "AlterEgo" regarding the linear response curve.... it's not needed for this capture to preserve highlight details.
Generally I also do use the linear response curve ... but not as my standard... only for certain captures when need is (I rarely shoot subjects with super high dynamic range, though). When you use the linear curve in base characterics you should be familar with using the curves tool. When you use the linear curve and only work with the contrast slider and the levels tool you may end up with "flat" and "dull" looking areas somewhere in your photos... sometimes things may look artificial (not film-like).

Thanks tho-mas.  Yes, the color noise reduction is way too high.  And I should have applied a tone curve with the Linear Response selected, so that should make a difference to the comparison.

One thing that I really don't like about C1 is the way that it leaves garbage behind in the image folders (specifically the .cop, .cot and .cof files in the Cache subdirectory).  Is there any way of moving the cache to a specific folder? Or failing that, to have the cache cleared automatically on exit? What I do at the moment is to run a batch file periodically to clean out the cache files, but it's a nuisance.

regards,

Robert

[AlterEgo]

Generally I also do use the linear response curve ...

I, myself, always use linear curve = but then I always postprocess /sometimes a lot/ in PS afterwards....

[Bart_van_der_Wolf]

I, myself, always use linear curve = but then I always postprocess /sometimes a lot/ in PS afterwards....

Yes, me too. When a Film curve response is baked into the output, the highlight details are compressed. That will make it harder (not impossible at 16-b/ch, but harder) to recreate subtle differences in postprocessing. While still processing in Capture One, there is little if anything lost, since all operations are parametric until rendered as output (display or file).

Since I would be postprocessing in Photoshop anyway, Clarity / Detail / deconvolution sharpening / etc., I prefer to start with a non-compressed tonecurve in highlights and shadows, and boosted midtone contrast. It especially makes shadows harder to recover if the micro contrast is already compressed (not as much as highlights, but still) before e.g. Clarity can access is.

When a subject needs decompression of certain tone ranges, I can do that with the Curves tool. It's often easy to see where the histogram steeply rises or falls, and at what levels there is a lot of tonality scattered in the image already. The control point picker also helps with addressing local issues or exploiting opportunities. It gives much more control than a generic film curve can do. And with adjustment layers one can apply multiple different adjustments locally.

The film-curve response does have another IMHO negative aspect to it, and that is that Exposure seems to be  almost a stop higher than it actually is in Raw or with a linear curve. That will lead to the need to pull exposure a lot, and apply highlight recovery which affects more than just the quasi 'clipped' highlights.

Cheers,
Bart

[AlterEgo]

Yes, me too. When a Film curve response is baked into the output, the highlight details are compressed. That will make it harder (not impossible at 16-b/ch, but harder) to recreate subtle differences in postprocessing. While still processing in Capture One, there is little if anything lost, since all operations are parametric until rendered as output (display or file).

Since I would be postprocessing in Photoshop anyway, Clarity / Detail / deconvolution sharpening / etc., I prefer to start with a non-compressed tonecurve in highlights and shadows, and boosted midtone contrast. It especially makes shadows harder to recover if the micro contrast is already compressed (not as much as highlights, but still) before e.g. Clarity can access is.

When a subject needs decompression of certain tone ranges, I can do that with the Curves tool. It's often easy to see where the histogram steeply rises or falls, and at what levels there is a lot of tonality scattered in the image already. The control point picker also helps with addressing local issues or exploiting opportunities. It gives much more control than a generic film curve can do. And with adjustment layers one can apply multiple different adjustments locally.

The film-curve response does have another IMHO negative aspect to it, and that is that Exposure seems to be  almost a stop higher than it actually is in Raw or with a linear curve. That will lead to the need to pull exposure a lot, and apply highlight recovery which affects more than just the quasi 'clipped' highlights.

Cheers,
Bart

yes, for a general use - but for a specific raw conversion comparison (in this topic, w/o further postprocessing, just ACR/LR vs C1, no PS) film curve works just fine (see the clouds on both sides - they are the same - http://postimg.org/image/veoqbqzqd/ ).... granted there are many ways to achieve the same result - with linear curve or film extra shadow, I personally found that standard film curve works here too... again not that I am using it myself ever or recommend to use in general.

[Jack Hogan]

The main objective of the tone curve is to make a typically 10+ stop DR image look good on a 6-8 stop output medium.  Here is what Adobe's standard curve looks like (from DNG Profile Editor):



If it looks good without it, don't use it because it will mess with chromaticities.

Jack

[Robert Ardill]

The main objective of the tone curve is to make a typically 10+ stop DR image look good on a 6-8 stop output medium.  Here is what Adobe's standard curve looks like (from DNG Profile Editor):



If it looks good without it, don't use it because it will mess with chromaticities.

Jack

Interesting ... I've created a dng profile (for use with LR) with a linear tone curve and it works very well.  The highlights clearly benefit, but so do the shadows.

So ... getting less disappointed with both C1 and LR

Robert

[Robert Ardill]

yes, for a general use - but for a specific raw conversion comparison (in this topic, w/o further postprocessing, just ACR/LR vs C1, no PS) film curve works just fine (see the clouds on both sides - they are the same - http://postimg.org/image/veoqbqzqd/ ).... granted there are many ways to achieve the same result - with linear curve or film extra shadow, I personally found that standard film curve works here too... again not that I am using it myself ever or recommend to use in general.

It's hard to see side-by-side, but the left image (linear one I guess) seems to have smoother transitions in the sky; and also less contrast.  The water is significantly darker and the image is less saturated.  It seems overall more natural to me.  I developed the image with both Film Standard and Linear and got pretty much identical results to yours (smoother skies, darker water, less saturated).  I found the same thing happens in Lightroom ... pulling back the highlights and bumping the shadows appears to increase the image saturation.  There is also less noise and more detail in the shadows under the bridge with the Linear curve.

Robert

[AlterEgo]

It's hard to see side-by-side, but the left image (linear one I guess) seems to have smoother transitions in the sky; and also less contrast.
 The water is significantly darker and the image is less saturated. 
It seems overall more natural to me.  I developed the image with both Film Standard and Linear and got pretty much identical results to yours (smoother skies, darker water, less saturated).  I found the same thing happens in Lightroom ... pulling back the highlights and bumping the shadows appears to increase the image saturation.  There is also less noise and more detail in the shadows under the bridge with the Linear curve.

Robert

no doubt that you like your own conversion more... because the left one as I clearly noted was from ACR using the conversion parameters recorded by you in your DNG  ... the point is however not to try to repeat ACR/LR conversions... if you want to switch totally then just try not to use  or think about ACR/LR at all for sometime

[Robert Ardill]

no doubt that you like your own conversion more... because the left one as I clearly noted was from ACR using the conversion parameters recorded by you in your DNG  ... the point is however not to try to repeat ACR/LR conversions... if you want to switch totally then just try not to use  or think about ACR/LR at all for sometime

Well, that's funny!  I guess that what is 'clearly noted' to you is not clearly understandable to me

In that case, I would say that the (film curve & C1) image on the right is giving a harsh result.  If you try the same image with a Linear curve you will see that it will be easier to process the image to be more muted and less saturated.

Robert

[tho_mas]

Yes, me too. When a Film curve response is baked into the output, the highlight details are compressed. That will make it harder (not impossible at 16-b/ch, but harder) to recreate subtle differences in postprocessing. While still processing in Capture One, there is little if anything lost, since all operations are parametric until rendered as output (display or file).

Since I would be postprocessing in Photoshop anyway, Clarity / Detail / deconvolution sharpening / etc., I prefer to start with a non-compressed tonecurve in highlights and shadows, and boosted midtone contrast. It especially makes shadows harder to recover if the micro contrast is already compressed (not as much as highlights, but still) before e.g. Clarity can access is.

The film-curve response does have another IMHO negative aspect to it, and that is that Exposure seems to be  almost a stop higher than it actually is in Raw or with a linear curve. That will lead to the need to pull exposure a lot, and apply highlight recovery which affects more than just the quasi 'clipped' highlights.

Bart,
while I basically agree I think it depends on the workflow. The "film curves" are designed to work in conjunction with a conventional lighting meter. But I'm sure I don't tell you any news :-)
Therefore when you shoot subjects with a high dynamic range and expose to the right the film curves provided in C1 may clip (and/or compress) highlights.
But when ETTR is not required so that you can roughly expose according to the lighting meter (to 18% grey so to say… or 1 stop above or so) while at the same time the dynamic range of the scene can be covered by your camera then the film curves may provide a great starting point for further editing.

Anyway… when you start editing with a linear curve you have to apply a curve sooner or later in the image editing process to gain a „natural“ or „filmic“ look. And I think the point in time when you apply a curve to create a certain look doesn’t really matter - provided you are editing in high bit (16bit) and the image content fits within the histogram (so no clipping in the shadows and highlights). Or to put it differently…: I think it doesn’t really matter whether you apply a curve in C1 or in later in Photoshop.
I for one do adjust the approximate look of my images in C1. I am a big fan of C1’s color editor and the whole thing only makes sense when you don’t alter anything later in Photoshop. I do, however, adjust my RAW files somewhat softer (contrast-wise) in C1 than the final image will look like to preserve „everything that has been captured“. Therefore all the subtile details that matter to me are there when I am processing to 16bit TIF (sharpening disabled on output, mostly with the camera profile embedded)… so no need to regain theoretically lost differentiation later in post.
To gain this I do use all the curves provided (except for the „high contrast“ curve) … depending on the image content and on how I’ve exposed the scene…

[ErikKaffehr]

Update: Raw image is here: http://echophoto.dnsalias.net/ekr/Articles/DRArticle/CanoeStadium/_DSC6262.ARW


Hi,

This is coming from my Alpha 99, not A7rII, but I would say it is an interesting case:

http://echophoto.dnsalias.net/ekr/Articles/DRArticle/CanoeStadium/20141109-_DSC6262.dng (I will post the ARW image this evening, the DNG has the ARW imbedded).

The image has very bright sky, with the disc of the sun recoverable, it also has dark detail. This is a typical high flare situtaion, however, so luminance ratio on sensor will be limited by lens flare.

This page points to some interesting details and has raw histograms and stuff.

http://echophoto.dnsalias.net/ekr/Articles/DRArticle/CanoeStadium/DS2.html

Regarding C1 processing, my take is that C1 sort of makes you underexpose, thus giving up on SNR (Signal Noise Ratio) and shadow detail to protect highlights. Capture One has pretty intensive noise suppression as default, probably to compensates for the low exposures.

To explain it a bit better, the film curve makes the image look much brighter than the linear curve, so an ETTR image will look overexposed at default settings. With LR ETTR images in contrasty situations look pretty good.

Best regards
Erik

[Robert Ardill]

... (if) the dynamic range of the scene can be covered by your camera then the film curves may provide a great starting point for further editing.

I agree. Also, I find that it is better to have the standard film curve as default for image selection (otherwise the images look too dark).  Then if a particular image has difficulties in the highlights or shadows we can go to a linear curve (which I do think helps a lot in these cases).

Anyway… when you start editing with a linear curve you have to apply a curve sooner or later in the image editing process to gain a „natural“ or „filmic“ look. And I think the point in time when you apply a curve to create a certain look doesn’t really matter - provided you are editing in high bit (16bit) and the image content fits within the histogram (so no clipping in the shadows and highlights). Or to put it differently…: I think it doesn’t really matter whether you apply a curve in C1 or in later in Photoshop.

That's something I've been wondering about.  Bart suggests that it would be better to apply Topaz Clarity (for instance) before applying the tone curve in the case of shadow recovery. I think you're saying that it should not matter once we keep to 16-bits.

Of course, applying a filter like Clarity on a linear image is going to be difficult to judge as the image will be rather dark, so from a visual editing point of view it would be better to apply the tone curve beforehand.

I for one do adjust the approximate look of my images in C1. I am a big fan of C1’s color editor and the whole thing only makes sense when you don’t alter anything later in Photoshop. I do, however, adjust my RAW files somewhat softer (contrast-wise) in C1 than the final image will look like to preserve „everything that has been captured“. Therefore all the subtile details that matter to me are there when I am processing to 16bit TIF (sharpening disabled on output, mostly with the camera profile embedded)… so no need to regain theoretically lost differentiation later in post.
To gain this I do use all the curves provided (except for the „high contrast“ curve) … depending on the image content and on how I’ve exposed the scene…

How important is it to edit colors as few times as possible?  In theory I would assume that the less the better, but even if everything is done in C1 or LR, presumably each of the adjustments are applied one after the other rather than as one combined adjustment?  In other words, let's say that there is a global saturation adjustment and a local saturation adjustment, presumably the global adjustment would be done first followed by the local adjustment.

If that's the case then it wouldn't seem to matter where the adjustments are made, providing we have a large enough color space and are working in 16 bits.

Just a question: do you think that using the embedded camera profile is a good idea?  The profiles are very small (around 200KB) which may result in interpolation errors (although I expect that P1 makes good profiles .. . and as the cameras are quite linear, small tables are probably OK).  I would have thought that if you want to keep as large a color space as possible that it would be better to go to a large working space like ProPhoto, rather than stay in the camera profile.

Cheers

Robert