Pages: 1 [2]   Go Down

Author Topic: Using both Capture One and Lightroom  (Read 10782 times)

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Using both Capture One and Lightroom
« Reply #20 on: April 16, 2016, 03:08:47 pm »

Hi Robert,

There are a lot of demosaic algorithms around, each has strengths and weaknesses. The problem is essentially that today's sensors undersample the image. Thus, there is a lot of detail that the sensor cannot resolve and this will be reproduced as fake detail. The way to solve that problem is to make the pixels smaller or to use an optical low pass filter that eliminates detail that the sensor cannot resolve. But, a correctly reproduced image will look very soft at actual pixels, so the images are normally under-filtered and have fake detail.

The enclosed image shows this quite clearly. The target at top is Bart's resolution test target while chart at the bottom is Norman Koren's MTF chart. Norman is the founder of Imatest if you would not know. The image on the left is Capture One and on he right is Lighroom CC-2015, both have a lot of issues although LR may be a bit worse than C1.

The image here was shot at optimum aperture with a good lens on a non OLP filtered sensor with relatively large pixels at 6.8 microns, that sensor is my P45+. You would not see the same amount of artefacts on the Sony A7rII, as it has much smaller pixels, 4.5 microns if I remember right. More importantly, the Sony sensor has gapless microlenses which means that it has much more of area sampling while the P45+ is closer to point sampling.

Each demosaic algoritm has it's own signature, check this article: http://www.libraw.org/articles/bayer-moire.html

I may guess that the halo artefacts come from LR overcompensating for the OLP filter on non OLP filtered images. I have not seen it on OLP filtered images.

Stopping down essentially eliminates those issue. On the P45+ I have it is necessary to stop down to f/16 to get rid of aliasing. In that case diffraction acts an OLP filter.

Best regards
Erik





Thanks Eric ... I'll have a good look when I get home tonight (I'm travelling at the moment).  It sounds interesting ... do you think the artifacts are removed by some converters, or created by poor demosaicing in others?

Cheers

Robert
Logged
Erik Kaffehr
 

Robert Ardill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 658
    • Images of Ireland
Re: Using both Capture One and Lightroom
« Reply #21 on: April 16, 2016, 04:25:07 pm »

Hi Robert,

You could check this for instance: http://forum.luminous-landscape.com/index.php?topic=104708.0

But also check this response: http://forum.luminous-landscape.com/index.php?topic=104708.msg860933#msg860933

Very interesting. I've downloaded your image (which is lovely) and developed in C1 and LR.  There is more stair-stepping visible with LR at 100% on screen, but then again LR gets the highlights back much better, so some of the artifacts in C1 are probably masked by being blown out.  Here is a crop of the 2nd plant with FM 1/200 applied (no sharpening or luminance noise reduction in either converter), then upsized by 2x using bicubic.  I don't know how valid this is as a test, but the C1 version on the left looks pretty ghastly compared to the LR version on the right.



Quote
With the A7rII (and the P45+) I often see extra contour artefacts like in the attachment below, check along the antenna and under the text 911. These are not ringing artefacts from sharpening but seem to arise in demosaic.
Erik

Very weird artifacts!!  They really look like sharpening artifacts ... perhaps LR is applying some base sharpening?

Regards

Robert
Logged
Those who cannot remember the past are condemned to repeat it. - George Santayana

Robert Ardill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 658
    • Images of Ireland
Re: Using both Capture One and Lightroom
« Reply #22 on: April 16, 2016, 04:50:17 pm »

Hi Robert,

There are a lot of demosaic algorithms around, each has strengths and weaknesses. The problem is essentially that today's sensors undersample the image. Thus, there is a lot of detail that the sensor cannot resolve and this will be reproduced as fake detail. The way to solve that problem is to make the pixels smaller or to use an optical low pass filter that eliminates detail that the sensor cannot resolve. But, a correctly reproduced image will look very soft at actual pixels, so the images are normally under-filtered and have fake detail.

The enclosed image shows this quite clearly. The target at top is Bart's resolution test target while chart at the bottom is Norman Koren's MTF chart. Norman is the founder of Imatest if you would not know. The image on the left is Capture One and on he right is Lighroom CC-2015, both have a lot of issues although LR may be a bit worse than C1.

The image here was shot at optimum aperture with a good lens on a non OLP filtered sensor with relatively large pixels at 6.8 microns, that sensor is my P45+. You would not see the same amount of artefacts on the Sony A7rII, as it has much smaller pixels, 4.5 microns if I remember right. More importantly, the Sony sensor has gapless microlenses which means that it has much more of area sampling while the P45+ is closer to point sampling.

Each demosaic algoritm has it's own signature, check this article: http://www.libraw.org/articles/bayer-moire.html

I may guess that the halo artefacts come from LR overcompensating for the OLP filter on non OLP filtered images. I have not seen it on OLP filtered images.

Stopping down essentially eliminates those issue. On the P45+ I have it is necessary to stop down to f/16 to get rid of aliasing. In that case diffraction acts an OLP filter.

Best regards
Erik

Thanks Eric ... so here it seems that C1 resolves down to around 80 while LR only goes to 70 (I guess 90 is Nyquist?).  After that we're into noise (or false image artifacts).  What's scary is what LR does between 70 and 100.

So what's the conclusion here? Is it that LR produces uglier rubbish than does C1 near Nyquist?  And perhaps that C1 resolves a bit closer to Nyquist?

Would applying noise reduction not go some way to fixing this issue, rather than stopping-down (at least to fix the issues with the demosaicing)? It does seem to help with your stair-stepping chapel :)

Cheers,

Robert
Logged
Those who cannot remember the past are condemned to repeat it. - George Santayana

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Using both Capture One and Lightroom
« Reply #23 on: April 16, 2016, 06:36:57 pm »

Hi Robert,

No, it is well known that it is essentially impossible to remove aliasing in post. The colour stuff can be reduced and some algorithms are more effective than others.

Nyquist should be around 73 lp/mm if the image was shot at the correct distance, around 5m. I guess the camera was a bit closer.

Now, this is an extreme case. The P45+ has large pixels and neither micro lenses or OLP filtering and the subject is very sensitive to aliasing. But pretty ugly aliasing can also occur on the A7rII. On the left the original image, on the right with "moire filter" on a brush. The "moire filter" smears out the colours.

The raw file for that image is here: http://echophoto.dnsalias.net/ekr/Articles/Shoots/Canon_16_35_4/20151211-_DSC3383.dng

Best regards
Erik





Thanks Eric ... so here it seems that C1 resolves down to around 80 while LR only goes to 70 (I guess 90 is Nyquist?).  After that we're into noise (or false image artifacts).  What's scary is what LR does between 70 and 100.

So what's the conclusion here? Is it that LR produces uglier rubbish than does C1 near Nyquist?  And perhaps that C1 resolves a bit closer to Nyquist?

Would applying noise reduction not go some way to fixing this issue, rather than stopping-down (at least to fix the issues with the demosaicing)? It does seem to help with your stair-stepping chapel :)

Cheers,

Robert
Logged
Erik Kaffehr
 

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Using both Capture One and Lightroom
« Reply #24 on: April 17, 2016, 02:59:23 am »

Hi Robert,

Thanks for liking my humble image. It was taken in the crypt of the Innichen Abbey in Innichen/San Candido.

I guess that the results we get from LR vs. C1 very much depends on processing skills, but I would also think that our experience with tools also dictates expectations. I am a very long time LR user, I started using it with the arrival of public beta 3, the first one that did run on the Windows platform I was using at that time.

I have played around with several raw converters aside from C1 and LR, like RawTherapee, Iridient's Raw Developer and AccuRaw. From the viewpoint of workflow I think LR is the one that suits me best. RawTherapee has a lot of advanced options and is designed by a lot of really smart people. RawTherapee has tons of options, like it is possible to select different demosaic algorithms. Amaze would be very good at reducing "zipper artefacts", while "LMSE" would be effective on moiré. All these programs support DNG and DNG Colour Profiles.

In LR it is possible to build HDR and panoramas that are still in DNG format, even if I think that DNG is half cooked. Those tools are not perfect but pretty good, in most cases.

Best regards
Erik


Very interesting. I've downloaded your image (which is lovely) and developed in C1 and LR.  There is more stair-stepping visible with LR at 100% on screen, but then again LR gets the highlights back much better, so some of the artifacts in C1 are probably masked by being blown out.  Here is a crop of the 2nd plant with FM 1/200 applied (no sharpening or luminance noise reduction in either converter), then upsized by 2x using bicubic.  I don't know how valid this is as a test, but the C1 version on the left looks pretty ghastly compared to the LR version on the right.



Very weird artifacts!!  They really look like sharpening artifacts ... perhaps LR is applying some base sharpening?

Regards

Robert
Logged
Erik Kaffehr
 

Robert Ardill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 658
    • Images of Ireland
Re: Using both Capture One and Lightroom
« Reply #25 on: April 17, 2016, 08:03:26 am »

Hi Robert,

No, it is well known that it is essentially impossible to remove aliasing in post. The colour stuff can be reduced and some algorithms are more effective than others.

Nyquist should be around 73 lp/mm if the image was shot at the correct distance, around 5m. I guess the camera was a bit closer.

Now, this is an extreme case. The P45+ has large pixels and neither micro lenses or OLP filtering and the subject is very sensitive to aliasing. But pretty ugly aliasing can also occur on the A7rII. On the left the original image, on the right with "moire filter" on a brush. The "moire filter" smears out the colours.

The raw file for that image is here: http://echophoto.dnsalias.net/ekr/Articles/Shoots/Canon_16_35_4/20151211-_DSC3383.dng

Best regards
Erik

I've checked out your image and C1 does a better job with the moire - less to start off with and removes it better than does LR (as far as I can see, at any rate).  But neither are too great ... have you tried the image with RT?

I checked out my image for artifacts, and there are some step-artifacts on the roof.  Again, C1 does a better job of keeping these down than does LR.  Here is a comparison, after over-strong FM sharpening of 2/100 to bring out the jaggies.  LR on the left:



Cheers,

Robert
Logged
Those who cannot remember the past are condemned to repeat it. - George Santayana

Robert Ardill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 658
    • Images of Ireland
Re: Using both Capture One and Lightroom
« Reply #26 on: April 17, 2016, 09:13:04 am »


I guess that the results we get from LR vs. C1 very much depends on processing skills, but I would also think that our experience with tools also dictates expectations. I am a very long time LR user, I started using it with the arrival of public beta 3, the first one that did run on the Windows platform I was using at that time.

I have played around with several raw converters aside from C1 and LR, like RawTherapee, Iridient's Raw Developer and AccuRaw. From the viewpoint of workflow I think LR is the one that suits me best. RawTherapee has a lot of advanced options and is designed by a lot of really smart people. RawTherapee has tons of options, like it is possible to select different demosaic algorithms. Amaze would be very good at reducing "zipper artefacts", while "LMSE" would be effective on moiré. All these programs support DNG and DNG Colour Profiles.


Hi Erik - I absolutely agree that from a workflow point of view LR suits me much better than the other raw converters I've tried.  I find RT almost unusable because it's so sluggish and complicated.  I do like C1 and would put up with the workflow inconvenience if I found that the increase in quality was significant ... but then I develop few pictures so a very fast workflow isn't important to me.  But I would still use LR for everything else, including image selection.

At this point though, I can see that C1 does some things better than LR and vice-versa: there's no clear winner for me. So I can't see myself using C1 except for images that I am struggling with in LR (and then I would give the image a go in C1 to see if it can do a better job for me).  As I always go to Photoshop for FM, Topaz etc (and of course for Photoshop itself), there's little that C1 can offer (me) unless it's superior demosaicing (both for detail and color).

I think you've demonstrated that C1 is better at reducing artifacts and Bart has shown that it renders fine details better (and I think the tests that I did with Imatest also show this).  So for large prints or for images with moire problems (and stair-step problems) it would probably be worth initial development in C1.

As for color ... I'm not sure.  For the images I've tried it's been the same: some images I preferred developed in C1 and others in LR.  But in general I've found that it isn't so hard to adjust the images to give an equivalent look.  But then I'm not in general looking for color reproduction (although I do photograph my paintings and paintings of friends and I have found that LR does a very good job, using the dng profiles).

Well, the good thing is that one learns by doing these tests :)

All the best,

Robert
Logged
Those who cannot remember the past are condemned to repeat it. - George Santayana

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Using both Capture One and Lightroom
« Reply #27 on: April 17, 2016, 08:09:19 pm »

Hi Robert,

Attachments below, with RT LMMSE, and Amaze. Also added "Accu Raw".

It was a good learning experience for me, too.

Best regards
Erik


Hi Erik - I absolutely agree that from a workflow point of view LR suits me much better than the other raw converters I've tried.  I find RT almost unusable because it's so sluggish and complicated.  I do like C1 and would put up with the workflow inconvenience if I found that the increase in quality was significant ... but then I develop few pictures so a very fast workflow isn't important to me.  But I would still use LR for everything else, including image selection.

At this point though, I can see that C1 does some things better than LR and vice-versa: there's no clear winner for me. So I can't see myself using C1 except for images that I am struggling with in LR (and then I would give the image a go in C1 to see if it can do a better job for me).  As I always go to Photoshop for FM, Topaz etc (and of course for Photoshop itself), there's little that C1 can offer (me) unless it's superior demosaicing (both for detail and color).

I think you've demonstrated that C1 is better at reducing artifacts and Bart has shown that it renders fine details better (and I think the tests that I did with Imatest also show this).  So for large prints or for images with moire problems (and stair-step problems) it would probably be worth initial development in C1.

As for color ... I'm not sure.  For the images I've tried it's been the same: some images I preferred developed in C1 and others in LR.  But in general I've found that it isn't so hard to adjust the images to give an equivalent look.  But then I'm not in general looking for color reproduction (although I do photograph my paintings and paintings of friends and I have found that LR does a very good job, using the dng profiles).

Well, the good thing is that one learns by doing these tests :)

All the best,

Robert
Logged
Erik Kaffehr
 
Pages: 1 [2]   Go Up