Old, dilapidated bridge

Quote:

---

Originally Posted by DanK

“HTH”?

---

"Hope This Helps", Dan.

Quote:

---

Is this approximated by frequency separation in Photoshop? I’ve only glanced at that but have been meaning to study it.

---

I've never used Photoshop, so I don't know what Adobe means by "Frequency Separation", sorry. Others here should be able to answer that ...

HTH ... ;)

The answer seems to be: sort of.

I've just started dabbling in frequency separation, spurred by this thread. however, both seem to be vaguely Fourier transforms, decomposing the image into components based on frequency, and they seem to have similar effects. Here is a very crude first try with frequency separation in photoshop:

https://photos.smugmug.com/photos/i-...LFBZ7gW-X2.jpg

One fundamental difference seems to be the application, not the math. As far as I have gotten with this in Photoshop, there appears to be no way to apply the transform automatically. Rather, one creates a layer (or several layers) that obscure high frequency detail using Gaussian blur, and then one merges the layers manually, using masks and opacity to fiddle with the application. In this case, I masked the entire original, with the high-frequency detail, and then painted it on with a brush over the bridge and foreground.

For anyone who is interested: I found several of the online tutorials about frequency separation confusing. A clear, basic one is here: https://photoshopcafe.com/frequency-...kin-photoshop/

Quote:

---

Originally Posted by DanK

The answer seems to be: sort of.

I've just started dabbling in frequency separation, spurred by this thread. however, both seem to be vaguely Fourier transforms, decomposing the image into components based on frequency, and they seem to have similar effects. Here is a very crude first try with frequency separation in photoshop:

https://photos.smugmug.com/photos/i-...LFBZ7gW-X2.jpg

One fundamental difference seems to be the application, not the math. As far as I have gotten with this in Photoshop, there appears to be no way to apply the transform automatically. Rather, one creates a layer (or several layers) that obscure high frequency detail using Gaussian blur, and then one merges the layers manually, using masks and opacity to fiddle with the application. In this case, I masked the entire original, with the high-frequency detail, and then painted it on with a brush over the bridge and foreground.

---

Looks pretty good, Dan. My try turned the whole image a bit fuzzy with perhaps too much of a "golden hour" look; yours, less so with more impression of depth given by the foreground highlights.

Living proof that there are many ways to skin a cat ...

What I did wasn’t quite frequency separation because I simply painted from one layer onto another rather than using the healing brush to pick up texture without color.

I'm intrigued by this technique now that I have tried it. I was thinking about a technique that might lessen the problem with the artificially blurred background behind Ole's piano player. That is, create several blur layers with different radii for the Gaussian blur--essentially, establishing different frequency ranges. Then paint them on one at a time with a very soft brush, starting with the smaller radius and working out to simulate greater blur at greater differences. One could also try multiple layers with the same blur but different opacity. I have no idea how well either would work, but it should be an interesting experiment once I have an appropriate image.

Quote:

---

Originally Posted by DanK

The answer seems to be: sort of.

I've just started dabbling in frequency separation, spurred by this thread. however, both seem to be vaguely Fourier transforms, decomposing the image into components based on frequency, and they seem to have similar effects. Here is a very crude first try with frequency separation in photoshop:

https://photos.smugmug.com/photos/i-...LFBZ7gW-X2.jpg

One fundamental difference seems to be the application, not the math. As far as I have gotten with this in Photoshop, there appears to be no way to apply the transform automatically. Rather, one creates a layer (or several layers) that obscure high frequency detail using Gaussian blur, and then one merges the layers manually, using masks and opacity to fiddle with the application. In this case, I masked the entire original, with the high-frequency detail, and then painted it on with a brush over the bridge and foreground.

For anyone who is interested: I found several of the online tutorials about frequency separation confusing. A clear, basic one is here: https://photoshopcafe.com/frequency-...kin-photoshop/

---

Frequency separation adds two layers, one containing clolour info, the other structure. So by using one or other of the layers, it's possible to clone colour without structure, and vice versa. It's often used in portraiture to correct skin blemishes, I use it mainly in infra red to correct hot spots, bright spots in the centre of the image caused mainly, I think, because the lenses are corrected for visible spectrum and not IR. I use Tony Kuyper's set of tools TK Actions, which has frequency separation as one of the tools.

Quote:

---

Originally Posted by DanK

<> For anyone who is interested: I found several of the online tutorials about frequency separation confusing. A clear, basic one is here: https://photoshopcafe.com/frequency-...kin-photoshop/

---

This one's also quite good, going more into how it works ...

https://fstoppers.com/post-productio...technique-8699

... and as a bonus, it has that composite of Einstein/Monroe a little way down ...

Hmm. I think the term "frequency separation" is slightly misleading.

leaving aside sharpening, the high-frequency image is the original. When you create a blurred copy, you have eliminated some high-frequency information--basically, microcontrast--but have retained lower-frequency data. However, but you have not removed color information at that point. You have muddled it a tiny bit. So, in the first stages of "frequency separation", you haven't really done much with color.

If I understand this, the distinction between detail and color that Derek mentioned--which is how frequency separation is described in some of the things I have read--arises because of the subtract blend mode. That should almost entirely remove color from the high-frequency layer--although not completely, as the blurring means that some pixels will have different colors in the two layers.

Am I off-base here?

Quote:

---

Originally Posted by DanK

Hmm. I think the term "frequency separation" is slightly misleading.

leaving aside sharpening, the high-frequency image is the original. When you create a blurred copy, you have eliminated some high-frequency information--basically, microcontrast--but have retained lower-frequency data. However, but you have not removed color information at that point. You have muddled it a tiny bit. So, in the first stages of "frequency separation", you haven't really done much with color.

If I understand this, the distinction between detail and color that Derek mentioned--which is how frequency separation is described in some of the things I have read--arises because of the subtract blend mode. That should almost entirely remove color from the high-frequency layer--although not completely, as the blurring means that some pixels will have different colors in the two layers.

Am I off-base here?

---

Can't say, Dan. I don't use layers, masks, brushes or blending for editing, and I know nothing about blend modes, sorry.