Pages: 1 [2]   Go Down

Author Topic: what's the best method to sharpen film scans without sharpening the grain?  (Read 1618 times)

BartvanderWolf

  • Sr. Member
  • ****
  • Online Online
  • Posts: 7411
Re: what's the best method to sharpen film scans without sharpening the grain?
« Reply #20 on: January 14, 2019, 04:19:23 PM »

What application do you have in mind that does "frequency separation"?

Hi Mark,

Affinity Photo has that functionality built straight into the application.

Official tutorial

Some user tutorials:
https://www.youtube.com/watch?v=3ZpVKaOKWOw
https://www.youtube.com/watch?v=C8_OXoPJ91c

Cheers,
Bart
« Last Edit: January 14, 2019, 04:51:10 PM by BartvanderWolf »
Logged
== If you do what you did, you'll get what you got. ==

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12292
    • http://www.markdsegal.com
Re: what's the best method to sharpen film scans without sharpening the grain?
« Reply #21 on: January 14, 2019, 04:21:00 PM »

Thanks Bart and Howard.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

kirkt

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 533
Re: what's the best method to sharpen film scans without sharpening the grain?
« Reply #22 on: January 14, 2019, 04:38:16 PM »

Here is an application-agnostic method for doing a single separation:

https://fstoppers.com/post-production/ultimate-guide-frequency-separation-technique-8699

It uses basic image editing tools so that it can be implemented in any image editor.  You can do a single separation or multiple frequencies if that is what you need.  The basic idea is to choose a blur radius that gets rid of your "high frequency" detail, which will include grain.  Then you can sharpen the residual image (the image without the high frequency) and composite the separation.  This is a classic technique used in portrait retouching, for example, where the artist wants to smooth or heal small blemishes and wrinkles (high- or middle frequency detail) on a separate layer while leaving the tonal sculpting of the face (low-frequency image data) intact.

As a variation, which will keep more real detail in the residual, low-frequency image, try using noise reduction software to perform the "blur" operation instead of a simple Gaussian blur.  Then use that result to generate the high frequency layer via Apply Image.  This will keep more high-frequency real detail in the residual, which is the image layer you want to sharpen (without the grain) and push the "noise" (grain) into the high-frequency layer that you will not sharpen, but composite back onto the sharpened residual layer via Linear Light blending mode.

You could also try Surface Blur instead of Gaussian Blur.

The scanning resolution and the size of the film grain relative to pixel size will dictate how completely or easily one can separate the grain content from the fine image detail.

kirk
« Last Edit: January 14, 2019, 05:47:17 PM by kirkt »
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12292
    • http://www.markdsegal.com
Re: what's the best method to sharpen film scans without sharpening the grain?
« Reply #24 on: January 14, 2019, 05:54:17 PM »

Thanks Kirk,

Reading this over, several observations come to mind. This procedure does not seem purpose-built for dealing with the problem of film "grain", which may have different characteristics from the subject matter of the retouching it appears to be purposed for, though not having done it, I don't have a view on how much of an issue this could be. It would appear to require case by case fine-tuning to work properly, which could pose a time conundrum for those doing large-volume scanning work. The distinction between image detail and so-called "grain" can be tricky, especially when dealing with naturally "grainy" surface textures. I can see the distinctions being difficult, yet that would be an essential element to successful separation. A bespoke application such as Neat Image allows one to easily identify and mitigate noise or "grain" by channel (Y - luminance; Cr, Cb - the colour channels) and by frequency (High, Medium, Low), allowing the user to decide in real time (as it recalculates) the level of adjustment for each at which image detail may begin to suffer. On top of the quality of the results, the ease of use of the interface and its responsiveness is what attracted to me to this program in the first place. While it may seem counter-intuitive to sharpen before "grain" mitigation, the advantage of working in this manner allows one to see in a post-sharpening context at what point the right degree of acutance of real edge detail that one achieved from sharpening could begin to suffer from overly aggressive "grain" mitigation in one or more channels or frequencies.

(You will notice I have "grain" in "", because when dealing with colour fiom it's a generic name for what may be so-called dye clouds that are formed by clumps of fundamental silver particles and dyes through nine layers of emulsion.) Dye-clouds are much larger than silver particles, not as sharply defined and come in random shapes, posing perhaps a challenge for distinguishing them from image detail and reducing their visibility without destroying real detail.)
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

kirkt

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 533
Re: what's the best method to sharpen film scans without sharpening the grain?
« Reply #25 on: January 14, 2019, 07:37:46 PM »

I use other software from these guys.

https://www.knowhowtransfer.com/photoshop-professional-plugins/wow-frequency-equalizer

I use that as well as several other panels from Davide and company. Terrific stuff.

Kirk
Logged

kirkt

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 533
Re: what's the best method to sharpen film scans without sharpening the grain?
« Reply #26 on: January 14, 2019, 07:40:59 PM »

Thanks Kirk,

Reading this over, several observations come to mind. This procedure does not seem purpose-built for dealing with the problem of film "grain", which may have different characteristics from the subject matter of the retouching it appears to be purposed for, though not having done it, I don't have a view on how much of an issue this could be. It would appear to require case by case fine-tuning to work properly, which could pose a time conundrum for those doing large-volume scanning work. The distinction between image detail and so-called "grain" can be tricky, especially when dealing with naturally "grainy" surface textures. I can see the distinctions being difficult, yet that would be an essential element to successful separation. A bespoke application such as Neat Image allows one to easily identify and mitigate noise or "grain" by channel (Y - luminance; Cr, Cb - the colour channels) and by frequency (High, Medium, Low), allowing the user to decide in real time (as it recalculates) the level of adjustment for each at which image detail may begin to suffer. On top of the quality of the results, the ease of use of the interface and its responsiveness is what attracted to me to this program in the first place. While it may seem counter-intuitive to sharpen before "grain" mitigation, the advantage of working in this manner allows one to see in a post-sharpening context at what point the right degree of acutance of real edge detail that one achieved from sharpening could begin to suffer from overly aggressive "grain" mitigation in one or more channels or frequencies.

(You will notice I have "grain" in "", because when dealing with colour fiom it's a generic name for what may be so-called dye clouds that are formed by clumps of fundamental silver particles and dyes through nine layers of emulsion.) Dye-clouds are much larger than silver particles, not as sharply defined and come in random shapes, posing perhaps a challenge for distinguishing them from image detail and reducing their visibility without destroying real detail.)

For purposes of this thread I experimented with using Neat Image to perform the “blur” operation for the frequency separation. It works well, given the level of control one can exercise over isolating the grain.

Kirk
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12292
    • http://www.markdsegal.com
Re: what's the best method to sharpen film scans without sharpening the grain?
« Reply #27 on: January 14, 2019, 09:02:28 PM »

Once you've used Neat Image, why do you still need frequency separation? Why not do the whole grain mitigation operation in Neat Image?
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

kirkt

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 533
Re: what's the best method to sharpen film scans without sharpening the grain?
« Reply #28 on: January 14, 2019, 10:24:55 PM »

Once you've used Neat Image, why do you still need frequency separation? Why not do the whole grain mitigation operation in Neat Image?

The original idea was to sharpen a film scan while not sharpening the grain present in the scan - I have assumed that the idea was not to completely remove the grain, but to somehow sharpen the underlying image and preserve the original grain.  Neat Image is good at isolating and removing the grain, which results in a grain-free "residual" image but with detail preserved (way better than using Gaussian Blur).  This low-frequency "residual" image is the part that you can sharpen without worrying about accentuating the grain in the original image.  The frequency separation technique essentially lets you subtract the grain from the original image and put it on its own layer - you will composite the grain back onto the sharpened residual image with the Linear Light blend mode.  This way you get to sharpen the image without the grain, but then add the original grain back in after the sharpening without the grain is performed.

All of this assumes that the grain is noise-like enough for Neat Image to be able to effectively isolate it.  Fortunately, Neat Image has enough control over noise type, channel and frequency that it is pretty adaptable for the task.  Also good - Neat Image can be set up to run in an Action.

Kirk
« Last Edit: January 14, 2019, 10:49:55 PM by kirkt »
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12292
    • http://www.markdsegal.com
Re: what's the best method to sharpen film scans without sharpening the grain?
« Reply #29 on: January 14, 2019, 10:45:36 PM »

Interesting approach, thanks.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

kirkt

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 533
Re: what's the best method to sharpen film scans without sharpening the grain?
« Reply #30 on: January 14, 2019, 11:35:43 PM »

You're welcome.

Here is an example of the difference between sharpening the grainy image with and without frequency separation as previously described.  This is an over-the-top example, but you get the point.  The original image is from a Fujifilm GFX 50s and grain was added in CaptureOne - is it really film-like?  I don't know, but this is for demo purposes.  The image shows a 100% crop of the original, the residual image resulting from treatment with Neat Image, the sharpened result using Smart Sharpen, radius = 2.0, amount = 500 applied to: 1) the residual image and recombined with the grain (left) and 2) the original, without frequency separation (right).  I had to reduce the composite image to 75% of its original size for purposes of making the JPEG easier to view and download, so some artifact occurred, but you can appreciate the difference between sharpening the detail and sharpening the detail and the grain.

kirk
« Last Edit: January 14, 2019, 11:42:06 PM by kirkt »
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12292
    • http://www.markdsegal.com
Re: what's the best method to sharpen film scans without sharpening the grain?
« Reply #31 on: January 15, 2019, 10:50:42 AM »

Thanks for doing this Kirk. Several queries:

(1) The image: not clear to me whether this is a good sample to work with. It has very little (nearly none) high density detail, so in a way, it doesn't need sharpening. Most of it is smooth surfaces with some distinct lines that don't obviously need much sharpening - I suppose depending on the magnification of the final photo. The advantage of this kind of image, however, is that it does show the effectiveness of grain mitigation on smooth surfaces very nicely. The disadvantage is that we have difficulty imagining impacts on the acuity of high density detail of the various approaches.

(2) The results: the top right version is the preferable outcome of course - looks fine, but did you notice that it has slightly toned-down highlights compared with the original? Not clear how that happened - whether a result of Neat Image, or something else. It's quite subtle but noticeable. Between the bottom two, even looking at them magnified, it isn't clear how much improvement the frequency separation brought to the right versus the left versions. There is some, but it's not that remarkable. Perhaps this is due to the limitations of the format in which such demos need to be posted here, I don't know.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."
Pages: 1 [2]   Go Up