K o r a y wrote:
We know how they do it. We just cant control that many variables at once and cant have a preview

??

If you're thinking of Nik's Tonal Contrast, you obviously found a very similar technique, but Topaz Detail works a bit differently.  IMO it's conceptually more akin to Lucis than to anything else, but I could always be wrong .

Also, you can get a 'preview' for the Median separation technique as follows:

1.) Duplicate the background / merged image x2.  We'll call the upper of these copies layer 1 and the lower layer 2.
2.) Invert layer 1.
3.) Set layer 1's blend mode to Linear Light, Fill 50%.
4.) Clip layer 1 to layer 2.
5.) Set layer 2's blend mode to Linear Light, Fill 50%.
6.) Run Median on layer 1.

You can also make layer 1 a smart object to have a refinable radius which you can work with as you go.  This isn't as 'accurate' a method as the traditional approach, but is within 1/256 values of it.

Edit: I may have incidentally come across the Tonal Contrast algorithm earlier this week.  If I get time this weekend, I'll try to implement it and see if we can't generate a free equivalent.

SRB Photo wrote:

??

I use surface blur not the median
The method I outlined here was the fastest way to explain things, mainly to give an idea about what one can possibly do.

The whole point was to control the detail even if its less scientific in my case. None of the plugins you mentioned above does something I cant get close to if needed.

Thats what I meant by power and detail is my bitch now

Lately what  impresses is the RawTherapee's quality shadow and highlights which I think is quite spectacular. Something nice is going on there

K o r a y wrote:
Lately what  impresses is the RawTherapee's quality shadow and highlights which I think is quite spectacular. Something nice is going on there

RT has a lot of features / implementations which are missing from its commercial cousins, and is a wonderful tool to have around.

SRB Retouch wrote:

RT has a lot of features / implementations which are missing from its commercial cousins, and is a wonderful tool to have around.

its a little unstable and stops working if you move  fast within

K o r a y wrote:

its a little unstable and stops working if you move  fast within

Fast is not its forte.  Try running an R-L deconvolution with 100+ iterations on a 21MP file.  You could take a nap with my system .

SRB Photo wrote:

Fast is not its forte.  Try running an R-L deconvolution with 100+ iterations on a 21MP file.  You could take a nap with my system .

Sean is it?

I have a question. First off thank you for bringing the SFS method to light. It's helped a ton.

But I was wondering if you or anyone has found a way to separate the color information from the GB layer as well?

I essentially would like 3 layers that combine to make the original image:

1) Detail Information
2) Straight Color Information
3) GB Monochrome Shaded Information

I'm not sure if this was touched on or not as this thread has garnered a lot of posts and they become a blur after a while.

DerW wrote:
To separate the color from the rest of the image the most precise solution I'm currently aware of would be to use a copy of the background layer set to "Color"-mode and place a monochrome layer in "Hue" mode beneath (e.g. background - layer filled with any shade of Gray in "Hue"-mode - copy of background in "Color"-mode).

Seems like there's no difference (just tested on one image in 8Bit so far).

Edit: And another image, in 16Bit, still no difference :-)

Hmm I just tried it and there seems to be a slight shift in contrast that happens. It's ever so subtle if you blink, but the noticed that the lips and hair of my subject ended up popping more lol.

I dunno about this lol.

Elite Retouch wrote:
But I was wondering if you or anyone has found a way to separate the color information from the GB layer as well?

I essentially would like 3 layers that combine to make the original image:

1) Detail Information
2) Straight Color Information
3) GB Monochrome Shaded Information

I'm not sure if this was touched on or not as this thread has garnered a lot of posts and they become a blur after a while.

It depends on what format you want that color information in.  It certainly can be done, and DerW's method is close to one of them - vs. using a copy of the original in 'Color' blend mode, simply use a duplicate of the GB layer in 'Color' blend mode - in quick testing, it's accurate to 1/32k.  There are other ways as well, including another Apply Image to pull off a color copy for Linear Light application, separating to LAB, using the HSB / HSL plugin, etc.  Like I said, though, how do you want the data formatted and what do you want to do with it?

SRB Retouch wrote:
It depends on what format you want that color information in.  It certainly can be done, and DerW's method is close to one of them - vs. using a copy of the original in 'Color' blend mode, simply use a duplicate of the GB layer in 'Color' blend mode - in quick testing, it's accurate to 1/32k.  There are other ways as well, including another Apply Image to pull off a color copy for Linear Light application, separating to LAB, using the HSB / HSL plugin, etc.  Like I said, though, how do you want the data formatted and what do you want to do with it?

Well I want it separate so I can even out the hues, but I also want to be able to manipulate the shading information without try to compensate for a color shift when values are changed.

I also want to see if this will give me more direct color of the saturation of some colors.

Elite Retouch wrote:
Well I want it separate so I can even out the hues, but I also want to be able to manipulate the shading information without try to compensate for a color shift when values are changed.

I also want to see if this will give me more direct color of the saturation of some colors.

Elite - I'm surprised that a retoucher of your caliber is asking these but better late than never right. Links below has some nice articles with some very important information:
http://www.thelightsright.com/view/tlr_tutorials_menu
http://goodlight.us/writing/tutorials.html
Take your time and enjoy the power.

Elite Retouch wrote:

Well I want it separate so I can even out the hues, but I also want to be able to manipulate the shading information without try to compensate for a color shift when values are changed.

I also want to see if this will give me more direct color of the saturation of some colors.

On the road but try the HSL / HSB plugin and work directly on the Hue & Sat, keeping the layer as a whole in Color blend mode.  If it snows really bad tomorrow Ill be able to put up an example.

Koray wrote:
Sean - do you know what Topaz refer to when they say "regions" ?

I haven't studied it in detail (ha!) but my gut is that it's similar to the # of "flat" areas formed by an operator similar to the median filter, surface blur, etc. It's also likely that that number involves furthe subdivisions, but again Ive not really analyzed it as yet.

SRB Photo wrote:

I haven't studied it in detail (ha!) but my gut is that it's similar to the # of "flat" areas formed by an operator similar to the median filter, surface blur, etc. It's also likely that that number involves furthe subdivisions, but again Ive not really analyzed it as yet.

You should.
Separation using dust and scratches or reduce noise results kind of their type of detail enhancing more or less versus surface blur.
They also strangely look more like deconvolution??

I cant really figure out the idea behind the regions which makes them stand out against all others.

Tez Mercer Photography wrote:
I've been following this thread from the beginning and admittedly, a lot of it flew right over my head.

Firstly, thanks for taking the time and effort to post this to help others out. I was told altruism was dead, but this reaffirms my belief in the opposite.

Anyways, I have the newer frequency action (the one with the curves layer included), and I use it this way:

Run the action,
Duplicate the blur layer at a much higher level (20-25 depending on the pic), this layer is underneath the 'linear light' layer.
I then add a layer mask to the linear light layer and paint with a black brush (about 10% opacity) to remove it so let the blurred layers come through, thereby softening the skin but not entirely removing the texture generated in the linear light layer.

Am i using this action in the right way? I feel a bit of a spoon leaving it so long to ask but if I just run the action as is, I get a hugely sharpened picture which is great for making people look old and craggy, but not the smoothing I was hearing about, so I did the layer mask thing, which seems to work, but I'm unsure if I'm missing out on something great by doing it.

Thanks again,

tez

Odd. If you're running the most current action, the result after separation should look identical to the original.

List your layer stack from top to bottom after running this. It might be something simple.

Tez Mercer Photography wrote:
Thanks for the reply, after running the action, the layers I have (from top to bottom) look like this:

Dodge/Burn (50% gray on softlight)

Sharpen (this is a group that contains HF1, 2 and 3 copies with masks)

HF Curves adj (used in relation to the below layer)

High Frequency (linear light)

Touchup- blank

Low Frequency

Frequency

-------------------------------------------------

What I do here is I lower the opacity of the linear light layer to make it look less embossed, dupe the low freq layer with a bigger gaussian blur, then create a mask for the areas i want to be softened. I then make a mask on the linear light layer to lower the opacity in certain parts for the smoothness to come through.

This does work really well in itself, but I don;t think this is what everyone else is doing and I get the impression I'm missing something fundamental.

Thanks again,

Tez

OK. That's the standard action set. Are you OK with editing actions? I can walk you through how to fix that problem. I thought I uploaded the newer one, but I guess I never did.

Edit: Nevermind. I uploaded the revised set. I'm not sure if it is in the same place as the one you got. Go here:

http://nunuvyer.biz/Photoshop/Separation/Actions/

And download the Frequency.atn file. That should do it for you. Sorry about that.

Tez Mercer Photography wrote:
ahh, thanks for that.

I downloaded, ran the action, but still find myself masking out the linear light layer to smooth the skin, or I dupe the LF layer and apply a median layer (which I then mask). I then bring back the texture of the skin by using a brush at 10% on the sharpen layer to make it less plasticky.

I have to admit that I feel like a luddite right about now.

Does the final image after separation look like the original? Check the settings of the curves, then. I didn't check that and it might be set to bump up contrast on the HF layer. In fact, that might work better for you and easier. Just adjust the curves layer to change the contrast and mask/unmask where needed.

K o r a y wrote:
Elite - What do you mean by the color pass & how do you make it?

Just asking to learn different approaches

Well the original method separates the image into 2 forms of information:

1) A fine detail layer with some tinting
2) A shaded hue layer (that GB layer)

I needed to gain ultimate control though because I know that sometimes when changing values, colors can get changed and vice versa. And it's such a painful process to then try to correct what you threw out of whack...So I know I needed:

1) A fine detail layer
2) A Hue layer (to control color variations and saturation)
3) A Shaded (Black & White) layer to control my values

So this is what I came up with (It's a little hard to see, sorry):



I made the color pass by duplicating my GB layer, changing the blend mode to color, and then placing a 50% gray layer underneath (in color blend mode) and merged those 2 layers together. I then desturated the original GB layer because I didn't need that color information anymore which then leaves me with JUST the value information.

Breaking the image up in this way allowed me to do all of my cloning/healing on the detail later, perform CC and hue tweaks on the color layer, and then do all of my D&B on the shading layer.

DerW informed me of how to make the color pass here: https://www.modelmayhem.com/po.php?thre … st12118449
(Be aware though that once you desaturate the GB layer, a slight change in value levels will occur. Nothing TOO major...but you will notice it.)

K o r a y wrote:
Got it thanks.
Do you do only one separation per image? If so how do you determine the initial radius.

Most of the time I do 3 separations, small, med and large.

This might sound kind of haphazard, but I only do 1. If it's a High Res image that I'm starting off with I just set the radius to 25 and I'm off to start working. And since we only accept High Resolution files...it works out fine.

I'm not sure if that radius is too much or too high for most of the people on here, but it's been working great so far for the last 8 retouches that I've done. Of course they've all been beauty/glamour/ headshots as well. I'll more than likely change the formula and make a couple different actions based on the size of the subject in the image (ie headshots, 3/4, full body...and so on).

Elite Retouch wrote:
(Be aware though that once you desaturate the GB layer, a slight change in value levels will occur. Nothing TOO major...but you will notice it.)

How exactly did you desaturate?

Saturation slider is the bad way.

NothingIsRealButTheGirl wrote:
How exactly did you desaturate?

Saturation slider is the bad way.

Umm...Image Menu> Adjustments> Desaturate

Please note that I'm saying the values of the image as a whole changes when using my 3 layer setup. I think it's something going on with the math the image layers are using. It's almost as if a local contrast is being applied without my say-so.

NothingIsRealButTheGirl wrote:
Saturation slider is the bad way.

In 'Normal' blend mode it does.  In 'Saturation' it does not (though still arguably just a bit imperfect).

For Elite, thanks for sharing this.  I want to write more, but it'll have to wait until later as I'm trying to work something else out atm .

Photo Realistic wrote:
The sharpening technique you described is called Difference of Gaussian and its a really good way to get higher degree of sharpening without the tell tale halo of over sharpening.

Another good technique is Difference of Surface Blur (DoSB). Same method but replace gaussian blur with Surface Blut and that will sharpen textures. Great for walls, fabrics and anything that has a lot of texture in it.

I could be wrong, but I'm rather sure the skin techniques here are largely based on DoGs (rather, inverses of them), but I don't believe that most of the sharpenings are.  If they were, they wouldn't increase the apparent noise (which they do).  If I'm wrong, though, I'll be happy to read how / why .

DoSB is also an interesting methodology, as is mixing GB and SB to create sharpening approaches.

Elite Retouch wrote:

Umm...Image Menu> Adjustments> Desaturate

Please note that I'm saying the values of the image as a whole changes when using my 3 layer setup. I think it's something going on with the math the image layers are using. It's almost as if a local contrast is being applied without my say-so.

If you use Image Menu> Adjustments> Desaturate on a bright seeming color like 100% yellow you will get the same gray as if you used it on a dark seeming color like 100% blue.

That could account for your value changes.

B Thomp wrote:
First off i want to thank everyone who put in any time and energy into this, i have only played around with it for a little bit but i am impressed by the possibilities.


I'm sorry if i missed this, i honestly didn't read everything because a lot of the beginnng was redone and redone...so i skipped to the end and read the last 4-5 pages...

There is a lot of talk about what this script can do....but i am at a loss to actually DO most of them. I am 100% self taught (no classes, but a lot of books and forum help) with PS and i know my way around pretty well but i am trying to figure out how to take this powerful tool and use it.

Can someone give a couple examples, a "how-to" on the top 3 things this can be used for.

I am no pro, and i wish nothing more then to learn from those who are MUCH MUCH smarter then i am...but this thread is hard to read. its a lot of grey area...not much black and white. I understand that i can play with all the variables to get different results....but can't we agree on a vanilla set of #'s and settings so i can start with vanilla and then work up to more complex things?

Right now it feels like i am trying to learn something much more complex then i can handle because i can't find the vanilla start i need to build from.


Many thanks to all of you guys/gals again. I am only giving my POV.

Photons 2 Pixels Images wrote:
If I get some time today (which I just might) I'll record a video of using separation to make skin cleanup a bit easier. Maybe I'll throw in another use or two.

OK. Things got hectic. I got a video done, though. This is a beginner's video so the rest of you don't look. I'd rather not be embarrassed.

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

Those who can't open .wmv files, I also put up a .flv file for Flash players...

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

I'm still figuring out all this video stuff so let me know what you'd like to see different.

P.S. If you watch this video, please at least let me know whether you got anything useful from it or not.

Tez Mercer Photography wrote:
thank you for the video.

I felt like I was fumbling around in the dark the first few times I ran the action, saw all the layers and was like "now what?" From reading various posts though, I pieced it together but never knew I was on the same track as everyone else or just flying blind.

Turns out, I was doing the 'right thing'.

One thing I do though is mask the high frequency layer and occasionally paint away the detail to smooth the skin (I mean I paint with a black brush on like 5% opacity). Not too much, as that defeats the point.

So, I found the video very helpful and reassuring in the way that I now know that I've been doing it right

I never used the retouch layer either, but now I can see its purpose. I had just been using color balance layers or hue/sat layers to color correct.

One of the things I like about the action is the dodge/burn layer. I find this really helps with outdoor work with models when you need to separate the background/foreground more, or to replicate Nik Color Efex's darken/lighten function to cast attention to one area of the photo.

I wouldn't change anything personally about the video. I found it easy to understand and coherent whilst watching you do your thang but maybe a PS newb might see it differently.

Thanks for doing that though, it's very much appreciated.

You're quite welcome.

One thing I'll use the Touchup layer for that I didn't mention in this video because it wasn't in the image....if you take a shot and there is a spot that has blown highlights, you can put some of the color/tone back in with this layer. Just another tool in the box.