Marcus J. Ranum wrote:
It's slightly OT but can someone describe for me what "apply image" does/means??  It's frustrating to me because I can see what it DOES but I lack the framework to understand it.

It's mixing the specified channels using the specified blending mode and opacity, and puts the results in the currently selected channel. (Or composite, if you've selected the RGB or CMYK 'channel'--which is rarely a good idea.)

It's kind of like the channel mixer, but you can only mix 2 channels at a time, while specifying different blend modes directly, as well as apply a mask to what is being applied, all in the same step.

Adobe's explanation isn't a bad place to start, though that covers using it in a very different way than most people use it. A websearch for 'photoshop calculations' will probably bring up some good guides--Calculations and Apply Image are very similar tools.

I tend to use Apply Image or Calculations when I want to build a mask from existing data, using the existing channels to create the mask for me. Or to mix data from a usable channel into an unusable one (blue channel in an underexposed shot taken indoors with Daylight setting, for example), so other salvage work can be done.

Sean Baker wrote:

I think I'm failing you here as I'm not sure what you're talking about.  I get CPD and that makes sense, but I'm lost at the spot per 100px?  Would it not be possible to simply do everything in terms of a pixel size, or were you referring to this in context of automating the skin workflow?

[Incidentally, re: the skin workflow, pupillary distance as measured with the measuring tool would be one reference which you could use to automate]

No, you aren't failing me. It isn't easy for me to explain what's in my brain sometimes, but I can see it clearly and it makes perfect sense to me. And the voices agree.

OK. My linoleum floor has something similar to a noise pattern on each square tile. By the "spot per 100 px" what I mean, to use the linoleum analogy, would be if I were to measure the amount of visible "noise" on a single tile (the area of) and count the number of noise "spots" then figure the average size of a spot with a margin of error.

I know it doesn't come out as actual "spots" but I couldn't think of another way to describe it. This would come out in terms of a pixel size. At least close enough to get a good reference for what you need based on the image you're working on.

Photons 2 Pixels Images wrote:
No, you aren't failing me. It isn't easy for me to explain what's in my brain sometimes, but I can see it clearly and it makes perfect sense to me. And the voices agree.

OK. My linoleum floor has something similar to a noise pattern on each square tile. By the "spot per 100 px" what I mean, to use the linoleum analogy, would be if I were to measure the amount of visible "noise" on a single tile (the area of) and count the number of noise "spots" then figure the average size of a spot with a margin of error.

I know it doesn't come out as actual "spots" but I couldn't think of another way to describe it. This would come out in terms of a pixel size. At least close enough to get a good reference for what you need based on the image you're working on.

I guess I'm struggling to understand how you mean to quantize said spots in an automated fashion?  It seems to me that if the goal is simply to provide replacement noise to a system wherein we've replaced a certain range that we can simply calculate the start and end sizes for the progressively scaled noise function which gives an identical result to an equivalently HP & GB'd copy.  Or was this for a different purpose in the first place?

Sean Baker wrote:

I guess I'm struggling to understand how you mean to quantize said spots in an automated fashion?  It seems to me that if the goal is simply to provide replacement noise to a system wherein we've replaced a certain range that we can simply calculate the start and end sizes for the progressively scaled noise function which gives an identical result to an equivalently HP & GB'd copy.  Or was this for a different purpose in the first place?

I think we're talking about the same thing but in different terms. Remember, I'm completely new to all this. It sounds to me like I'm thinking the same thing you're saying, I just can't get it out right.

And you are correct, this would be to calculate the range that we are trying to achieve in order to replace what was removed. The result I'm thinking in my head would be consistent with the GB radius used in the separation.

WOW! It's too early for me to be thinking about this without my coffee in me yet. I hope I'm making sense.

Photons 2 Pixels Images wrote:
I think we're talking about the same thing but in different terms. Remember, I'm completely new to all this. It sounds to me like I'm thinking the same thing you're saying, I just can't get it out right.

And you are correct, this would be to calculate the range that we are trying to achieve in order to replace what was removed. The result I'm thinking in my head would be consistent with the GB radius used in the separation.

WOW! It's too early for me to be thinking about this without my coffee in me yet. I hope I'm making sense.

Put like that I agree we're talking similarly with different terms used.  Coffee is just beginning to flow here, so I'm a bit behind.  I think there's potential but it's going to require a number of attempts at implementation with 'field' validation in PS to verify that the indended frequencies are being output by the algorithm.  A huge step forward if it works, though.

-----

And on a separate note, I've been thinking on the difference between pixel-level D&B and blur based techniques and what makes them fundamentally so different.  Each is designed to locally level contrast at varying sizes (radii for blur; brush size for D&B), so why do we think of them as fundamentally unlinked?  And what I'm wondering, perhaps due to personal bias, or simply lack of caffeination, is whether it has to do with replacement texture.  Again this is where I think bias may be creeping in, but work with me for a second - no matter how good the retoucher, no one traces perfectly with the brush through the crevice / pore / etc. to reshape without creating a bit of an edge where one didn't exist previously.  And isn't that the same as creating new texture at the same spatial frequency as the brush size?  Reducing that frequency in one location, but recreating a bit of it elsewhere?  Selectively replacing a frequency range with a more even application in the same range?  I may not be explaining that well; it's only just occurred to me; but for anyone interested and reading, I'd love to hear some thoughts.

Smedley Whiplash wrote:
Wishful thinking...

I wish there was a 'color spatializer' action that does for color, what this action does for sharpening. 

What do you want to do with the color that you're trying to separate?

Sean Baker wrote:

Put like that I agree we're talking similarly with different terms used.  Coffee is just beginning to flow here, so I'm a bit behind.  I think there's potential but it's going to require a number of attempts at implementation with 'field' validation in PS to verify that the indended frequencies are being output by the algorithm.  A huge step forward if it works, though.

-----

And on a separate note, I've been thinking on the difference between pixel-level D&B and blur based techniques and what makes them fundamentally so different.  Each is designed to locally level contrast at varying sizes (radii for blur; brush size for D&B), so why do we think of them as fundamentally unlinked?  And what I'm wondering, perhaps due to personal bias, or simply lack of caffeination, is whether it has to do with replacement texture.  Again this is where I think bias may be creeping in, but work with me for a second - no matter how good the retoucher, no one traces perfectly with the brush through the crevice / pore / etc. to reshape without creating a bit of an edge where one didn't exist previously.  And isn't that the same as creating new texture at the same spatial frequency as the brush size?  Reducing that frequency in one location, but recreating a bit of it elsewhere?  Selectively replacing a frequency range with a more even application in the same range?  I may not be explaining that well; it's only just occurred to me; but for anyone interested and reading, I'd love to hear some thoughts.

I think I get what you're saying. I'm at work on my cellphone so can't give this my full attention but I think you're definitely on to something. This is kinda what I'm hoping to accomplish. Once the unwanted frequencies are removed they can be replaced with something more visually pleasing yet still natural looking without going pixel level. It would save a lot of time for sure.

Smedley Whiplash wrote:
I wish there was a 'color spatializer' action that does for color, what this action does for sharpening. 

Which part? Hue or saturation?

You can create saturation masks (though it's a different kind of a pain than tone frequency masks), and you can make hue-based masks as well.

Sean Baker wrote:
Quick note re: texture / noise replacement of removed frequency bands.  I've not worked out the why yet, but I recently stumbled across the fact that applying a highlight mask to the replacement texture while blending it in Soft Light gives a much, much more realistic and pleasing effect.  I suspect it has something to do with captured vs. artificially-injected tonal information but again that's a subject for another day.  The point is, it's (IMO) eerily realistic @ ~20% opacity.

I'm an @sshat.  It's because Soft Light is protecting the highlights.  I still think it looks better than a lowered-contrast linear light blend, but the blend mode is the reason.

Kevin_Connery wrote:

Which part? Hue or saturation?

You can create saturation masks (though it's a different kind of a pain than tone frequency masks), and you can make hue-based masks as well.

Sean Baker wrote:
What do you want to do with the color that you're trying to separate?

I guess I'm wishing there was some action that would separate warm tones from cool tones (within the same image), or even separate by some user defined perimeter.

Ah, after looking at CS4, I see that they've simplified masks, so maybe I just need to upgrade. I know I can do it manually in CS3, but it takes a long time.

My goal is to simplify the color palette,  so the photo takes on a look like old school  illustrated posters-



See how the palette in this example is different from normal photography? I just thought that maybe there might be a way to spatially separate color like we are doing with tone frequency masks.

grahamsz wrote:
I think i've come up with something even more interesting. Here's a step by step tutorial on how I applied a selective bandstop filter to eliminate skin noise with periods between 5 and 15 px.

Here's the original image:



I then use photons 2 pixels action above to perform spatial frequency separation using a 5px blur. Once that's done I move to the low frequency layer and lasso the skin, taking care to avoid the eyes, lips and hair.


Now I apply a 15px feather, followed by a 15px gaussian blur. This in effect eliminates a particular range of frequencies in the skin, which in signal processing should make it bandstop filter.


And finally i switch to the high frequency layer and clone out blemishes


Total editing time for this little sample, about 3 mins.

Apologies again to the model here. She's actually got great skin, I just have a bunch of shots of her with fairly hash sunlight hitting her skin at an unflattering angle. I really should go take some of myself and use those.

i am not sure what i am doing wrong, but i am not liking it, I have tried this and i got people saying its too blurry, and i agree

http://i18.photobucket.com/albums/b130/ … 1248845296

Craziest Benny wrote:
i am not sure what i am doing wrong, but i am not liking it, I have tried this and i got people saying its too blurry, and i agree

http://i18.photobucket.com/albums/b130/ … 1248845296

If you click on all of the frequency layers so they are active, the image will be extremely sharp, then you can drag back opacity on each layer to taste. That's what's so awesome about it.

I've found that the HF3 layer is the most critical in terms of any blur factor, and I usually leave that one at 100%, while HF2 & HF3 are better somewhere between 25-45%.

The copy layer masks (HF1-3 copy), can be painted in to give you an extreme amount of additional sharpness, or just a little if your brush opacity is set low.

The amount of variance available is astounding.

Smedley Whiplash wrote:

If you click on all of the frequency layers so they are active, the image will be extremely sharp, then you can drag back opacity on each layer to taste. That's what's so awesome about it.

I've found that the HF3 layer is the most critical in terms of any blur factor, and I usually leave that one at 100%, while HF2 & HF3 are better somewhere between 25-45%.

The copy layer masks (HF1-3 copy), can be painted in to give you an extreme amount of additional sharpness, or just a little if your brush opacity is set low.

The amount of variance available is astounding.

hmmm thank you for this tip, i will try it out soon

Craziest Benny wrote:

hmmm thank you for this tip, i will try it out soon

Just curious...which action set are you using for this?

Virtuoso Skins wrote:
Photon, I'm going to test out your actions soon, been meaning to for a while, but have been so busy.

I notice a lot of fr separation for editing. One quick way I do it, to eliminate ISO noise or isolate your colors while editing details, is simply go into LAB color and work the channels.

I have been experimenting fairly heavily on various methods of enhanced sharpening trying to utilize the luma channel in LAB color, but so far I have not figured out anything earth shaking that surpasses anything discussed here.

I was using a High Pass filter on the Lum channel from LAB mode to sharpen edges but found that I could only go so far with it before getting halos. Then I'd ether have to back off to a point where I didn't think it was sharp enough or go through a pixel at a time to mask out the halos. Then Sean came along with the frequency separation. I haven't looked back since. There is still some artifacting and halo noticeable when you push it too far, but not nearly as much as the High Pass method.

If you don't work in 16 bit mode, make sure you switch over to it first before running the action or grab the script from the first action set I posted here. I should include the script with all of them, actually.

And by all means, if you can think of anything else you want to see in there let me know.

Photons 2 Pixels Images wrote:

I was using a High Pass filter on the Lum channel from LAB mode to sharpen edges but found that I could only go so far with it before getting halos. Then I'd ether have to back off to a point where I didn't think it was sharp enough or go through a pixel at a time to mask out the halos. Then Sean came along with the frequency separation. I haven't looked back since. There is still some artifacting and halo noticeable when you push it too far, but not nearly as much as the High Pass method.

If you don't work in 16 bit mode, make sure you switch over to it first before running the action or grab the script from the first action set I posted here. I should include the script with all of them, actually.

And by all means, if you can think of anything else you want to see in there let me know.

I will soon as I get my main machine up that has CS4, been in a tech hell, thought my MB died, got a new one, rebuilt the machine, it's not the MB Back to ordering parts

So I'm stuck with CS for a while longer, so I can't test out your action because of the smart shapes

I do want to see how it compares to Lukos, which works well for me on some images.

Virtuoso Skins wrote:

I will soon as I get my main machine up that has CS4, been in a tech hell, thought my MB died, got a new one, rebuilt the machine, it's not the MB Back to ordering parts

So I'm stuck with CS for a while longer, so I can't test out your action because of the smart shapes

I do want to see how it compares to Lukos, which works well for me on some images.

OK. Let me start on an action set that will work with CS. I just have to work around the Smart Object part. Should be easy enough.

Photons 2 Pixels Images wrote:
OK. I updated the original Frequency.zip file with the new action set that is backward compatible to Photoshop 7 (when Apply Image was introduced.)

http://www.nunuvyer.biz/Photoshop/Frequency.zip

Some things to note in this one. There are 2 Action Groups along with the script included in this one. The new stuff is in the Action Group "Frequency.atn" with the Photoshop 7 action labeled as Photoshop7 once you load this set into Photoshop.

If you are using the script due to working in non RGB/16 mode, make sure this Action Group is named "Frequency" and the action you want to run is named "Layers" or change the parts in the script on line 21 to reflect what you have.

Any problems just let me know. I'm home today and other than some yardwork, I'll be bored.

OK. I didn't have CS to test this one. It seems CS doesn't have some of the features I tried to include in the action set so it won't work. Sorry. I'll see what I can do to work around that.

Photons 2 Pixels Images wrote:

OK. I didn't have CS to test this one. It seems CS doesn't have some of the features I tried to include in the action set so it won't work. Sorry. I'll see what I can do to work around that.

Good news....maybe. I installed Photoshop 8 (CS) so I can test the scripts/actions to make sure they are backward compatible and if not, make adjustments. So, gimme some time and I'll figure out where the version I just posted went wrong.

Cycle_xpressions wrote:
WOW, I have been following this post for a while and with my limited skills I have been afraid to attempt it, but I gave it a whirl tonight and it looks amazing.

One question, and again remember I am really poor at photoshop, but when I do blemish editing on the LF layer it is really blurry so I am unable to see the edits.  I must be missing something right?

Also the tones layer, what am I missing here?  I have tried to play with this layer and I am not seeing the results, what is the function of this layer and how is it best used?

Also on the workflow side of things, is everyone making color adjustments then doing this action?  I was just curious what everyone's workflow is in using the action.

Thank you guys so much, there are a few major players that made this action happen but there are also alot of people who suggested great things to refine the process.  You guys are amazing and my hats are off to you for being secure enough in your craft to share with us.

Thanks to everyone

You aren't missing anything here. When the action finishes running (depending on which action you are using) the image should look the same as the original or, if you grabbed the last action (Others.atn or Others.zip) then it will appear a bit sharper.

For skin corrections, you use the LF layer to adjust general skin tones to even them out some. You use the HF layer to do the spot touchups such as cloning, healing, etc.

I hope that makes sense.

Jinxt Photo wrote:
I am also waiting in anticipation for Koray's explanation on how he got the effect produced in the SB example of his GB/SB/Nik comparison post.

Its so easy that I decided not to make it any easier...Here is how I do it:

Detail Passing: by Koray

- Instead of duplicating the background layer I switch to channels and control click the RGB channel and duplicate that instead in channels palette. Why?...I dont know, I'm following Seans instructions but somewhere else. So duplicate whatever you want and experiment.
You just have to have two copies of whatever. There are no rules.

- Now instead of gaussian, I apply a surface blur. watch the edges and adjust the threshold accordingly and set your radius higher than the threshold. Since I dont know what I'm doing I dont have any numbers or fancy ratios for you

- Then you invert the blurred channel and apply the other using the add command like Sean says no questions asked.

- If things are all right, you should be looking at a channel with some ghostly details and once you see it, you will know its not a highpass effect.

- Then you select all, copy, and paste it on top of your top layer in layers palette.
Set its blending mode to linear light as its been instructed or experiment with others but the effect will be visible when:

- You adjust the contrast and everything else of your Detail Map. You can do everything you want but start with a clipped adjustment I recommend.
I tried things like blurring and sharpening it, run a shadow and highlights etc. its fun.
And its actually very powerful. Combine it with a regular blur separation and things get way too real and sharp and detailed

Here is the same copy of the comparison image:


I'm being unclear on purpose. Ask when you get stuck.

Plus I think I made up two new ps terms

syd47421 wrote:
Syd's Graphic Equalizer

needs a couple of tweaks in the levels outputs, but the thing works.. and you can use the theory of it to make as many separations as you need.

I almost hate to ask but my brain is getting to tired to search.  How do I load in a script such as the one above?  I looked around in CS4 and I already knew where my scripts but didn't see anything that jumped out at me telling me how to load a new one.

Thanks

Viewu wrote:

I almost hate to ask but my brain is getting to tired to search.  How do I load in a script such as the one above?  I looked around in CS4 and I already knew where my scripts but didn't see anything that jumped out at me telling me how to load a new one.

Thanks

Copy the script to your "Presets/Scripts" folder within the Adobe/Photoshop folder.

Then restart Photoshop.

If done right, you should now have access to it under File>Scripts>whatever

Photons 2 Pixels Images wrote:

Copy the script to your "Presets/Scripts" folder within the Adobe/Photoshop folder.

Then restart Photoshop.

If done right, you should now have access to it under File>Scripts>whatever

Thanks!

Koray wrote:
I'm being unclear on purpose. Ask when you get stuck.

Plus I think I made up two new ps terms

So far so good. I'm seeing how the Surface Blur creates more of a detail map than a High Pass effect. Matching up a SB layer alongside the typical BG plus some clipped level adjustments creates a friggin scary-ass detailed and sharpened image. It's almost like LucisArt or TopazAdjust without ANY of the cartoonish halos. This must be what deconvolution is all about.

I've been toying with it on full-res 12mp files and I'm liking the effect when I run SB at 30/15 and 40/20. It keeps all of the defined edges where they need to be and I end up pulling plenty of detail out of it. It IS very much reminding me of Nik's Tonal Contrast

This makes me want to test what happens with other types of blurring just for the hell of it.

I'm not quite understanding why a Luminosity Mask channel is being created unless it is being used for masking some of the adjustments and such (I didn't use it for anything so I may be doing it different than you are).

This is something I'm definitely going to spend some time with.

Viewu wrote:
Thanks!

I'm sorry, I wasn't paying attention. My bad.

You are trying to load an action, not a script. To load an action, go to your actions palette and in the upper right corner, there is a drop-down menu. Click that, then click Load Actions...

Navigate to where the action is and load it. You may want to place it in the Actions folder first and load it from there.

To be on the safe side, once you do this restart Photoshop and make sure your action is still there.