I Must Be Dead wrote:
I read about 15 pages of this and my head just hurts. Not sure I even learned anything. I think I'm confused since the technique spread out into different categories like sharpening and skin smoothing.

And I would regard myself as very good with photoshop lol. So I perhaps am just confused on what is being done to each layer to sharpen, or smooth skin or what have you.

Perhaps I'm confused because I'm a visual learner and would prefer to see someone doing it in a youtube video. Anyone wanna do that!

The intro post explains how to lay it out, but to me is a bit blurry when explaining what what to do to which layers after you have got your low-frequency and high-frequency layers. Further it seems now if someone is confused then the people who are not so confused are just telling us to go back and read again. I apologize if it is obvious or makes sense, but I would love to see this layed out a bit more simple and dumbed down once again even though it has been, or a video made to show exactly how to sharpen or how to smooth skin using the technique.

I seriously felt like I was switching back and forth between what people were trying to do with the technique every page or so haha.

Anyhow, I'm very interested and curious. And I've been reading for three hours. But have not figured anything out!

3 hours? Try 14 months.

And I still haven't figured it out.  But it's glorious.

DerW wrote:
Just the page before this one gives you two videos: http://www.model-citizens.com/Tutorials … ion-1.flv, http://www.youtube.com/watch?v=kcY0Anrd5Yk (maybe this one as well: http://www.youtube.com/watch?v=2D2GZgUs9BQ although not the best example) and here's another one: http://www.digitalphotoshopretouching.c … ir-retouch

If you've still got some questions left, feel free to ask :-).

Best regards,
Jonas

The first link has a added comma in the link http://www.model-citizens.com/Tutorials … tion-1.flv

http://www.model-citizens.com/Tutorials … tion-1.wmv

I am a huge fan man retouching and restoring images are made so much simpler once you understand simply where and how to retouch the different layers. I started using a surface blur on the color and low pass layer and love the results.


Here's a similar tutorial from retouchpro:

Chain;271603 wrote:
It's out!

Here it is; with actions for download:
Skin retouching (Norwegian)
Skin retouching (google translate > english)

SRB Photo wrote:
More on this when I finish the article proper, but to dovetail with what Jonas has said above (he is correct that the above does not get 100% results - very close, but not quite):

One can get 100% reconstruction in 8bpc or 16bpc by using three layers to do so (vs. 2).  I'll explain the math in the article, but here's the procedure:

1.) Duplicate the image to separate 3x (twice if you don't care to retain a copy of the original for comparison).  For discussion, we will now call these the bottom, middle, and top layers.
2.) Blur the bottom layer as you see fit - for testing, do something intermediate to make sure you get enough that you should see loss if it's present.
3.) Disable the top layer; select the middle layer.
4.) Apply Image.  Select the bottom layer as Source, RGB as channel, check the Invert box, choose 'Add' as your blend mode, 100% opacity, Scale 1, Offset 0.
5.) Disable the middle layer; select and show the top layer.
6.) Apply Image.  Select the bottom layer as Source, RGB as channel, do not check the Invert box, choose 'Subtract' as your blend mode, 100% opacity, Scale 1, Offset 0.
7.) Show all layers.
8.) Set the top layer's blend  mode to Linear Dodge.
9.) Set the middle layer's blend mode to Linear Burn.
10.) If desired, compare the merged result to the original - you should find 0/0/0 discrepancy.

But like I said - more later.

So I get the advantage of splitting the image in two layers but what's the advantage of splitting it in three? A more targeted band pass filter, sharpening?

Ruben Vasquez wrote:

So I get the advantage of splitting the image in two layers but what's the advantage of splitting it in three? A more targeted band pass filter, sharpening?

Pixel-for-pixel reconstruction of the original image.  For most people and most purposes it wouldn't be a meaningful difference vs. splitting to two layers, but for the forensic community and the truly anal it might make a difference.  For the rest, it's just an interesting bit of understanding about how PS works to know.

FLEXmanta wrote:
Freq split tutorial by me on hiendworkshops. Sorry for compression, i delivered HD. Also, it's hard tu put it all in a

SRB Photo wrote:

Pixel-for-pixel reconstruction of the original image.  For most people and most purposes it wouldn't be a meaningful difference vs. splitting to two layers, but for the forensic community and the truly anal it might make a difference.  For the rest, it's just an interesting bit of understanding about how PS works to know.

Ok. So for general frequency seperation type moves, two layers is fine then?

Ruben Vasquez wrote:
Ok. So for general frequency seperation type moves, two layers is fine then?

Of course. Split, work, merge... split again later, work some more, merge again. You can choose a different radius every time and as many times you want. That's why the high pass sucked, cause you coudn't, and now you can because the split is identical to the merged. The split is one of those techniques that works well when you work on a "pixels on the bottom, adjustments on top" basis.

Ruben Vasquez wrote:
Ok. So for general frequency seperation type moves, two layers is fine then?

For 99.9% of work anyone is going to do, absolutely.  I don't believe in doing 2 layers for 8bpc editing, but I accept that some do and still get results which are acceptable for them.  OTOH, for those working in 16bpc, using >2 layers for a single-frequency split would, IMO, be over the top.

As FLEXmantra states, the old HP technique would destroy high contrast detail, whereas this technique loses 1 level in a fairly distributed fashion which isn't eye-catching (and, in 16bpc, visible on any monitor I know of).

Ruben Vasquez wrote:
Is there a way to do a frequency seperation in LAB? I don't seem to be having much luck when I attempt it.

It can be done - it's just a PITA for the way PS handles LAB blend modes and its Brightness / Contrast implementation.

Do everything the same as using the techniques which allow use of the HP filter, except, instead of using the Brightness / Contrast adjustment to halve contrast (it doesn't work in LAB), fill the layer with 50% gray, and then Fade that fill by 50%.

It's just a different (and more involved) way of halving the image's contrast, but one which plays nice with LAB.  From there, you can use the HP filter like we have before and your reconstruction is good to ~Fill.
7.) Choose 50% Gray.  Click OK.
8.) Edit->Fade.
9.) Fade by 50%, blend mode Normal.  Click OK.
10.) Change High Frequency layer to Linear Light blend mode.
11.) Run High Pass filter at same radius as in (4).
12.) Done!

picturepipeline wrote:
Hi @all.

I have been using this method for a while now and am really happy with the results. I do have two questions though and am hoping somebody can help

Sometimes when editing the high layer close to edges (for example removing stray blond hairs close to the hairline in shadow areas) I get color smears. How can this be explained and what am I doing wrong? I am using a hard edged brush and my understanding is that color smears shouldn't appear anyway because there is no color information on the high level. I've went through it over and over again, following the instructions here carefully. Anybody have an idea what is going wrong? Second question: what influence does the chosen radius for gaussian blur (low level) have on apply image (high level)?

Any help appreciated. Thanks in advance.

Greetings from Germany
Andy

I'm guessing you're talking about using the healing brush near edges.

First, the HF layer will retain color information. I have yet to do a separation when this did not occur. The nature of the beast.

Second, the smears you're talking about I'm guessing are from use of the healing tool. The healing tool samples from the surrounding area for tone and from your sample source for detail then tries to figure out what you want. If you're near an edge, it will pull some information from the edge itself and incorporate that into the result. There are a few ways to get around this. If the edge is fairly straight, take your sample with the edge centered on your sample point. Then when you start healing, place the cursor so the edge of the area you are healing is centered on the cursor. If you have the overlay turned on, you can line up the sampled edge with the target edge. Then do your healing.

Another way is to use the clone tool near an edge instead of the healing brush. Just remember that you only want to sample current layer.

If I'm misunderstanding you, forgive me. But is this more or less where your problem is coming from?

Pixels 2 Pixels wrote:
I'm guessing you're talking about using the healing brush near edges.

First, the HF layer will retain color information. I have yet to do a separation when this did not occur. The nature of the beast.

Second, the smears you're talking about I'm guessing are from use of the healing tool. The healing tool samples from the surrounding area for tone and from your sample source for detail then tries to figure out what you want. If you're near an edge, it will pull some information from the edge itself and incorporate that into the result. There are a few ways to get around this. If the edge is fairly straight, take your sample with the edge centered on your sample point. Then when you start healing, place the cursor so the edge of the area you are healing is centered on the cursor. If you have the overlay turned on, you can line up the sampled edge with the target edge. Then do your healing.

Another way is to use the clone tool near an edge instead of the healing brush. Just remember that you only want to sample current layer.

If I'm misunderstanding you, forgive me. But is this more or less where your problem is coming from?

Good points, mentioned above.

In addition, you can also make a selection of the area close to the edge and use the heeling brush. Heal in inside the selection. The selection will limit the effect so that it doesn't sample the area near the edge and smudge it all into one undesired stroke. I usually use the Pen tool for the selection.

Krunoslav-Stifter wrote:

Good points, mentioned above.

In addition, you can also make a selection of the area close to the edge and use the heeling brush. Heal in inside the selection. The selection will limit the effect so that it doesn't sample the area near the edge and smudge it all into one undesired stroke. I usually use the Pen tool for the selection.

I've tried the selection thing and it still pulled information from the edge (outside the selection). I'll have to look into that again.

Pixels 2 Pixels wrote:
I've tried the selection thing and it still pulled information from the edge (outside the selection). I'll have to look into that again.

It's a tip that you can actually find in Photoshop official help manual.

"If there is a strong contrast at the edges of the area you want to heal, make a selection before you use the Healing Brush tool. The selection should be bigger than the area you want to heal but should precisely follow the boundary of contrasting pixels. When you paint with the Healing Brush tool, the selection prevents colors from bleeding in from the outside. "

Press CTRL+H to hide the selection temporarily and you will be able to see the effect better, without any marching ants acting as a distraction. Just don't forget to press CTRL+H again, when you are done. lol. It's like CAPS LOCK and the brushes.

Also I forgot to mention, that if you use Replace blend mode the healing brush will act like the clone tool and you can use it to paint in near the edges and than blend the outer edge with healing brush set to normal blend mode. (Also from the manual)

"Mode - Specifies the blending mode. Choose Replace to preserve noise, film grain, and texture at the edges of the brush stroke when using a soft‑edge brush."

It applies ti hard-edge brush as well, they just didn't though about it.

Another interesting fact.
"Aligned - Samples pixels continuously, without losing the current sampling point, even if you release the mouse button. Deselect Aligned to continue to use the sampled pixels from the initial sampling point each time you stop and resume painting."

http://help.adobe.com/en_US/Photoshop/1 … 7607a.html

Pixels 2 Pixels wrote:

I've tried the selection thing and it still pulled information from the edge (outside the selection). I'll have to look into that again.

I have experienced the same bug using a selection maybe ithe selection was feathered don't know always switched to cloning or changed heal brush to replace but a pain since I forget to switch it back and wonder why it's not working right lol....

SRB Photo wrote:

It can be done - it's just a PITA for the way PS handles LAB blend modes and its Brightness / Contrast implementation.

Do everything the same as using the techniques which allow use of the HP filter, except, instead of using the Brightness / Contrast adjustment to halve contrast (it doesn't work in LAB), fill the layer with 50% gray, and then Fade that fill by 50%.

It's just a different (and more involved) way of halving the image's contrast, but one which plays nice with LAB.  From there, you can use the HP filter like we have before and your reconstruction is good to ~Fill.
7.) Choose 50% Gray.  Click OK.
8.) Edit->Fade.
9.) Fade by 50%, blend mode Normal.  Click OK.
10.) Change High Frequency layer to Linear Light blend mode.
11.) Run High Pass filter at same radius as in (4).
12.) Done!

Thanks Sean! Iv been gone for awhile so I'll have to play with this for sure. I'm a little suprised by the methods though. I would have thought the technique would focus on the L channel only. Do you still have the advantage that LAB offers, that of seperating color and contrast?

Ruben Vasquez wrote:
Thanks Sean! Iv been gone for awhile so I'll have to play with this for sure. I'm a little suprised by the methods though. I would have thought the technique would focus on the L channel only. Do you still have the advantage that LAB offers, that of seperating color and contrast?

You can do it that way as well - applying the 50% gray fill to the L* channel alone and fading.

I just don't personally like separating color from the frequency separation process [at least not with images of people].

SRB Photo wrote:

You can do it that way as well - applying the 50% gray fill to the L* channel alone and fading.

I just don't personally like separating color from the frequency separation process [at least not with images of people].

Any particular reason why?

Ruben Vasquez wrote:
Any particular reason why?

The color and luminosity / lightness interact in life to create images which look natural to us.  The same is true when you do your frequency separations.  There will be hue shifts during the transition from highlight to shadow, saturation differences between these and between midtones, etc.  By completely separating color from luminosity, I find that the result often has color inconsistencies on close inspection - places where those natural variations have become inconsistent with their corresponding luminous pixels.  Certainly there may be situations where it's advantageous to do so - I've just never found it to be so when working with people.

SRB Photo wrote:

The color and luminosity / lightness interact in life to create images which look natural to us.  The same is true when you do your frequency separations.  There will be hue shifts during the transition from highlight to shadow, saturation differences between these and between midtones, etc.  By completely separating color from luminosity, I find that the result often has color inconsistencies on close inspection - places where those natural variations have become inconsistent with their corresponding luminous pixels.  Certainly there may be situations where it's advantageous to do so - I've just never found it to be so when working with people.

I can think of a couple. Typically lips are much more positve in the A channel then the rest of the skin which allows some wiggle room for ajusting saturation of just the skin while maintaining saturation or adjusting to taste, the saturation of the lips. Also dark circles under the eyes are typically less saturated then the rest of the skin which is easy to adjust in LAB without affecting the rest of the face or the need of masks.

DerW wrote:
I'm just reading the thread over at RetouchPRO and thought I might throw in an idea or two of mine as well :-).

The apply image technique is nothing I would change at all, it works great!
The B/C idea however I'm still not sure about (I'll come to that in a minute).
Instead of using the B/C (applied directly) to split the frequencies (or for that matter create a low pass and a band stop in the next step), I'm thinking about using 50% fill opacity instead.
Using the fill has the benefit that some values aren't clipped that much (as you might know, fill calculates the blending with the correct data, instead of opacity, which uses clipped data), so in my tests (I checked with the histogram) the fill was a little more accurate than the B/C (very little though).

Maybe it's worth checking and I'd be very interested to see what you think about it :-).

Best regards,
Jonas

Jonas - how are you applying the 50% fill to create the effect?  Inverted, blurred copy overtop in LL with 50%F, or something else?  Thanks.

I tested this pretty thoroughly before, so I'll be surprised to find any differences this time around (it terms of quality).  Just the same, I'm always open to new ways of skinning the cat.

DerW wrote:
Edit: As I said, I'm not sure about this, so I thought you might want to test it as well. Didn't know you already did ;-)

Can you tell me how you generated your B/C layer?  I've looked at it and agree that the version you're showing has a considerable difference from 'right' due to clipping, but when I create the B/C version myself get replication within 1/32k points in all channels.  Let me know!

DerW wrote:
Hehe, I wasn't reading the RP-thread while I was testing here, so sorry about that ;-)

Best regards,
Jonas

How could I blame you for not reading that thread?  My head hurts trying to deal with the fact inoculation which seems to have gone on in there.