Forums > Digital Art and Retouching > HighPass Sucks (+ solution)

Photographer

Sean Baker Photo

Posts: 8044

San Antonio, Texas, US

Herman van Gestel wrote:
what do you actually mean with clipped curve?

1. Starting with the spatially-separated image.
https://www.twicebakedphoto.com/download/Curves/Step1.jpg

2. With the high frequency layer selected, choose Image -> Adjustments -> Curves (or from the layers palette while holding down Alt).  EDIT: Charles Yeh has quite rightly pointed out that I meant Layer->New Adjustment Layer->Curves here.  My bad.
https://www.twicebakedphoto.com/download/Curves/Step2.jpg

3. Check the "Use Previous Layer to Create Clipping Mask" box.
https://www.twicebakedphoto.com/download/Curves/Step3.jpg

4. The key to getting more traditional (sharpening / blurring) results from this is to make sure that, however you shape your curve, you keep an anchor point at 128 / 128 to preserve your middle grey.
https://www.twicebakedphoto.com/download/Curves/Step4.jpg

Hope that's a better explanation guys.

Apr 21 09 03:21 am Link

Photographer

JeF Briguet

Posts: 119

Leytron, Valais, Switzerland

Hello Sean,
Thank you for sharing this!
I'm using almost the same procedure (clipping mask after a GB/HP) to reduce moire in jeans. I never thought about using it for sharpening.
For prints/publication i'm heavily using incredimental HP on skin (male subject) and after some testings it looks your method is even better at this.

Tips for moire/softening: try an inverted U-curve on the clipping mask.

Apr 21 09 03:59 am Link

Photographer

Click Hamilton

Posts: 36555

San Diego, California, US

[comment deleted]


I started to make a wise crack to one of Bob Randall's wise cracks, then decided I'm over my head and would rather contribute something more positive to this thread.

It's time to zip my lips and pay attention.




Sean, you got the cement mixer between my ears revolving again.

Especially about the importance of tools, learning curves, and time, relative to results. And about how we might apply this knowledge for other purposes.

With luck, something good might start flowing down the chute.


Best wishes from San Diego

Apr 21 09 04:57 am Link

Photographer

Luminis Media Arts

Posts: 55

Atlanta, Georgia, US

This is awesome, many, many thanks, I've already replaced this as a part of my workflow, and opened my eyes to a lot of other possibilities with the apply filter tool. 

I'm amazed that somone finds it necessary to comment that they think it's useless because it's too technical.  I would just think "ok, I'll come back when I learn more."   Just be glad, if everyone was able to use all the tools PS had to offer, there'd be a lot more competition than there already is.

Apr 21 09 07:07 am Link

Photographer

Sean Baker Photo

Posts: 8044

San Antonio, Texas, US

JeF Briguet wrote:
Tips for moire/softening: try an inverted U-curve on the clipping mask.

Oh that's just neat.  Like masking the highlights out but more fun.  Also makes me realize how painfully aware of how restricted I've been in my usage of curves.  Thank you!

Apr 21 09 10:42 am Link

Photographer

Michael Magers

Posts: 4050

Fullerton, California, US

When I first did this all I got was a grey layer. Granted I may have been missing a step??  Dunno but while playing around with it I deleted the "Blur" layer copy and it came up like the HP but with not all the noise.  Not sure if you are supposed to keep that layer but after the mask has been done delete the blur layer and check your results.

Apr 21 09 11:27 am Link

Photographer

Sean Baker Photo

Posts: 8044

San Antonio, Texas, US

AidanJamesphotography wrote:
When I first did this all I got was a grey layer. Granted I may have been missing a step??  Dunno but while playing around with it I deleted the "Blur" layer copy and it came up like the HP but with not all the noise.  Not sure if you are supposed to keep that layer but after the mask has been done delete the blur layer and check your results.

It sounds like either you selected the wrong layer in the Apply Image dialog, or (more likely) that the blend mode on the high frequency data was still set to 'Normal'.  I think I understand your last line to mean that you did in fact get the sharpening effect you wanted in the end?

Apr 21 09 12:15 pm Link

Photographer

denisemc

Posts: 555

Toronto, Ontario, Canada

I knew I'd been missing out not being familiar with the apply image tool. Thanks so much!

Apr 21 09 12:26 pm Link

Photographer

Michael Magers

Posts: 4050

Fullerton, California, US

Sean Baker wrote:

It sounds like either you selected the wrong layer in the Apply Image dialog, or (more likely) that the blend mode on the high frequency data was still set to 'Normal'.  I think I understand your last line to mean that you did in fact get the sharpening effect you wanted in the end?

Yes I did get the effect in the end.  I am running CS2 and I went thru the steps two or three times.  Even did the 16bit one to see if I was doing it wrong.   Actually I think I did miss something but don't remember what it was. But my results, I did not GBlur very much maybe 4.5, was not that noticeable.  But when I took that layer away that's when I saw a significant change.

Apr 21 09 12:28 pm Link

Photographer

Robert Randall

Posts: 13890

Chicago, Illinois, US

Click Hamilton wrote:
[comment deleted]


I started to make a wise crack to one of Bob Randall's wise cracks, then decided I'm over my head and would rather contribute something more positive to this thread.

I usually crack wise on the ner do wells and nay sayers in a thread like this to keep them quiet or out completely. I figure if I embarrass them enough, the thread might just stand a chance of helping someone. I know I learned two big things from it, and that doesn't happen very much anymore. Its not that there isn't a lot I can learn, its just really hard to find someone that isn't simply re-packaging a Scott Kelbyism.

Keep it coming!

Apr 21 09 12:33 pm Link

Photographer

denisemc

Posts: 555

Toronto, Ontario, Canada

AidanJamesphotography wrote:

Yes I did get the effect in the end.  I am running CS2 and I went thru the steps two or three times.  Even did the 16bit one to see if I was doing it wrong.   Actually I think I did miss something but don't remember what it was. But my results, I did not GBlur very much maybe 4.5, was not that noticeable.  But when I took that layer away that's when I saw a significant change.

The same thing happened to me so I just reduced the opacity of my blur layer really low because I thought the effect was a little strong. Not sure what I missed, but I really like the effect and think I achieved it anyway.

Apr 21 09 12:35 pm Link

Photographer

Sean Baker Photo

Posts: 8044

San Antonio, Texas, US

AidanJamesphotography wrote:
But when I took that layer away that's when I saw a significant change.

That's the way this is designed to work.  That is, you still have copies of both the high and low frequency data - with your choice to do HP sharpening, add a little bit of Orton-ness to your picture, or something even crazier.  All at your fingertips.

Edit to add: The steps outlined are not designed to give you any sharpening in the end (though obviously a 100% sharpen is one click away).  Your image should look just the same as it did to start; only now it is on two layers with two very different types of information in them for you to utilize.  I'll update the OP to make this a bit more clear, and I apologize for having led any of you astray.

Apr 21 09 12:35 pm Link

Photographer

Black Russian Studio

Posts: 1431

New York, New York, US

Sean Baker wrote:

1. Starting with the spatially-separated image.
https://www.twicebakedphoto.com/download/Curves/Step1.jpg

2. With the high frequency layer selected, choose Image -> Adjustments -> Curves (or from the layers palette while holding down Alt).
https://www.twicebakedphoto.com/download/Curves/Step2.jpg

3. Check the "Use Previous Layer to Create Clipping Mask" box.
https://www.twicebakedphoto.com/download/Curves/Step3.jpg

4. The key to getting more traditional (sharpening / blurring) results from this is to make sure that, however you shape your curve, you keep an anchor point at 128 / 128 to preserve your middle grey.
https://www.twicebakedphoto.com/download/Curves/Step4.jpg

Hope that's a better explanation guys.

Thanks Sean. It does improve yout recipie, and eliminates need for duplicating sharpening layer if 100% was not enough, plus adds extra precise contol over sharpening.

Apr 21 09 04:19 pm Link

Photographer

Yingwah Productions

Posts: 1557

New York, New York, US

Sean Baker wrote:

You can use your same way of starting this process.  Setup your channel mixer adjustment layer in your image.  Now select the uppermost visible layer and choose Stamp Visible to create a merged copy of the entire scene by pressing Shift+Ctrl+Alt+E (Cmd+Shift+Alt+E on Mac).  With the new layer selected, run the action as before.  You've now separated your created channel's spatial frequencies and can use either the high or low data to suit your needs, exactly as you did in the past, mind you a bit more accurately.  As Mr. Kevin notes, it will not be as dramatic out of the box as the old HP method if you use it for sharpening / local contrast - the advantage lies in having fewer artifacts.

I tried adding in the channel mixer layer. Interestingly it upped the saturation alot, kinda made a cool effect. That was fixed by converting the resulting layer to greyscale again.
Adding in the channel mixer reduced some of the halo-ing effect of your method. But I think you can do the same thing with your curve adjustment layer. I'll have to try again with a good headshot to see how it affects the finer details.

Apr 21 09 08:12 pm Link

Photographer

Mikael Ramirez

Posts: 154

Tucson, Arizona, US

Just wanted to say thanks to OP for this awesome method. I love it.

-Mikael

Apr 22 09 12:31 am Link

Photographer

MEK Photography

Posts: 6571

Westminster, Maryland, US

Pretty sweet Sean!

I tried this out last night, and I'm liking it. 

First one in my port is using this.  (18+ if you want to check it out.)

Apr 22 09 04:44 am Link

Photographer

KEKnight

Posts: 1876

Cumming, Georgia, US

WOW!!!!  Great stuff Sean.  Thanks for posting!!  smile

Apr 22 09 10:08 am Link

Photographer

Hector Fernandez

Posts: 1152

Mexico City, Distrito Federal, Mexico

Havent had the time to try this till this morning. Wow my files are so sharp that it can cut my monitor. Anyway thanks and if ever in Mexico City come to my studio for free beers.

Apr 22 09 10:16 am Link

Photographer

Joshua Nykamp

Posts: 187

Denver, Colorado, US

works really well.

Apr 22 09 10:53 am Link

Photographer

Shutterslam

Posts: 224

Sudbury, Massachusetts, US

Just adding to the pile.

I had to read your post a few times before the penny dropped.  I was lucky to have even come across the high pass sharpening technique to be able to appreciate that there was a more accurate way of building that sharpening mask.

I'm not fully understanding a few points.

* I work on 8 bit jpegs, should I convert to 16 bit?  The workflow for 8 bit seems to work, but I'm not 100% sure.

* for the Gaussian Blur step, what would the "ideal range of values"  be for say a 12mp photo (4200 x 3000x) (i.e. 4, 10, 20 etc...)

* if I understand correctly, to apply the sharpening effect at the end of the workflow, you can
a) apply a curves adjustment with anchor points at 128 128
b) use smart sharpen on the spatial frequency layer
c) use unsharp mask on the SFL?

I can see how the method introduces sharpening with fewer artifacts, so I think I'm applying this method correctly, I just want to make sure I understand how the values are being chosen.

Thank you for your very valuable contribution to this community.

Lito

Apr 22 09 01:39 pm Link

Photographer

Gibson Photo Art

Posts: 7990

Phoenix, Arizona, US

I ran both your method Sean and a HP filter and see a massive difference between the two.

Now the hard part. Trying to figure out how to use it properly in my workflow. smile

Apr 22 09 06:43 pm Link

Photographer

Sean Baker Photo

Posts: 8044

San Antonio, Texas, US

Robert Randall wrote:
simply re-packaging a Scott Kelbyism.

I suspect that I'd have been recommending the latest and greatest (sponsored) plugin were that the case big_smile.

Yingwah Productions wrote:
I tried adding in the channel mixer layer. Interestingly it upped the saturation alot, kinda made a cool effect. That was fixed by converting the resulting layer to greyscale again.
Adding in the channel mixer reduced some of the halo-ing effect of your method. But I think you can do the same thing with your curve adjustment layer. I'll have to try again with a good headshot to see how it affects the finer details.

Were you checking the 'Monochrome' box when doing this?  I realize I neglected to include suggestion for doing so in my previous post (sorry) and wonder if that's the difference?

How were your headshot results?

Hector Fernandez wrote:
Havent had the time to try this till this morning. Wow my files are so sharp that it can cut my monitor. Anyway thanks and if ever in Mexico City come to my studio for free beers.

Absolutely big_smile.

Shutterslam wrote:
* I work on 8 bit jpegs, should I convert to 16 bit?  The workflow for 8 bit seems to work, but I'm not 100% sure.

Use whichever workflow works for you; each version of the technique is tailored to give the best result for that workflow which I could come up with.  If you note a degradation in quality, let me know and I'll reexamine, but they should be about as accurate as PS will allow.

Shutterslam wrote:
* for the Gaussian Blur step, what would the "ideal range of values"  be for say a 12mp photo (4200 x 3000x) (i.e. 4, 10, 20 etc...)

That is a matter of taste, and depends on subject content, capture sharpening (if applied), detail in the image, and your output method & size.  I can't give you a single number which will be best for anything you do.  I'll tell you that when I work with an 800x530 image, I tend to use ~.6 or .8.  When I work at full resolution, it's usually scaled from that (so for a 4200px image I would first try 3.6), but I always tune it to the image.  There are other folks here, though, who can speak to output sharpening far better than I - printing is still something which is very new for me.

Shutterslam wrote:
* if I understand correctly, to apply the sharpening effect at the end of the workflow, you can
a) apply a curves adjustment with anchor points at 128 128
b) use smart sharpen on the spatial frequency layer
c) use unsharp mask on the SFL?

Any of these are valid ways of applying additional sharpening, and can even be combined so long as you do so carefully (particularly while using SS or USM).

Shutterslam wrote:
Thank you for your very valuable contribution to this community.
Lito

You're absolutely welcome, and feel free to continue asking questions as well as sharing your own findings.

Apr 23 09 08:22 am Link

Photographer

Robert Randall

Posts: 13890

Chicago, Illinois, US

Robert Randall wrote:
simply re-packaging a Scott Kelbyism.

Sean Baker wrote:
I suspect that I'd have been recommending the latest and greatest (sponsored) plugin were that the case big_smile.

Hey... you took that our of context... put it back!

smile

Apr 23 09 08:32 am Link

Photographer

Sean Baker Photo

Posts: 8044

San Antonio, Texas, US

Robert Randall wrote:

Hey... you took that our of context... put it back!

smile

I'm sorry.  I guess you meant that as if were just posting the "secrets" to the Dave Hill technique? roll smile

Edit: And for anyone I've led astray in posting that link, please look here for a better run down.

Apr 23 09 08:40 am Link

Photographer

Yingwah Productions

Posts: 1557

New York, New York, US

Sean Baker wrote:

Were you checking the 'Monochrome' box when doing this?  I realize I neglected to include suggestion for doing so in my previous post (sorry) and wonder if that's the difference?

How were your headshot results?

Where's the monochrome box?

Edit: nevermind you meant in the channel mixer
yes its checked, but it wouldn't matter because its still a rgb image and its reading information from the image below in the apply image step(unless i have that concept wrong)

I also posted incorrectly, if you do the channel mixer on the bottom layer it ups the saturation, if you do it to the top layer it desats(useful for that dave hill look which I personally dislike). If I do the mixer on both layers then yes it works normal.

havn't had a chance to try it on headshot, been actually working. shocking!

Apr 23 09 11:20 am Link

Photographer

BT Imagery

Posts: 1020

Christiansted, Saint Croix, Virgin Islands of the United States

Nevermind, problems.quote]

Apr 23 09 02:35 pm Link

Photographer

Sean Baker Photo

Posts: 8044

San Antonio, Texas, US

Yingwah Productions wrote:
havn't had a chance to try it on headshot, been actually working. shocking!

Psh, paid work!  Showoff! lol

Biggs Photography wrote:
Here is an updated action that combines the layers from the original action into the original document.  Let me know if someone has problems with the download or action.

Download

Not seeing a difference in the downloaded file, but I'd love to know how you did it smile.

Apr 23 09 07:26 pm Link

Photographer

Sean Baker Photo

Posts: 8044

San Antonio, Texas, US

Has anyone bothered to use a healing brush on the high frequency detail of a model's skin yet?

Apr 23 09 07:36 pm Link

Photographer

BT Imagery

Posts: 1020

Christiansted, Saint Croix, Virgin Islands of the United States

Biggs Photography wrote:
Here is an updated action that combines the layers from the original action into the original document.  Let me know if someone has problems with the download or action.

Not seeing a difference in the downloaded file, but I'd love to know how you did it smile.

Sorry, I had a moment. The link should take you to the appropriate and working action now.

P.S. Thank you for doing all this, it will help many of our workflows in the long run I think.

Apr 24 09 05:31 am Link

Photographer

Robert McCadden

Posts: 171

Kenmore, Washington, US

Great stuff Sean; I just used this for the first time last night and was amazed at how sharp and clear I was able to get things.  The shot I used this for is the headshot that is currently my avatar in case anyone is interested in seeing how this looks on a headshot.

Apr 24 09 09:09 am Link

Photographer

LeDeux Art

Posts: 50123

San Ramon, California, US

Sean Baker wrote:

Sure.  The image below is the difference between a HP + GB solution for frequency separation and the original image.  Processing was done in 16bit mode with a 5px radius when the image was 800x527.  Resized to save reading space, but no other modifications were made.

https://www.twicebakedphotography.com/download/Demo.jpg

The same difference image is pure black when separated using the above technique.  I can post it if you like, but it really is all black in 8bit (JPG).

Let if that's what you were hoping to see.

Sean

very interesting, thank you for sharing

Apr 24 09 09:12 am Link

Photographer

Jose Deida

Posts: 1293

Reading, Pennsylvania, US

Crazy control! It's playtime...Thank you!

Apr 24 09 09:45 am Link

Photographer

Robb Mann

Posts: 12327

Baltimore, Maryland, US

Sean,

You've moved past photoshop. Check out this package instead: http://www.ittvis.com/

It's designed to do the kinds of things you're looking for. Raw, total control of every pixel.

Sadly, it's not designed for photography, but rather remote sensing.

Apr 24 09 09:53 am Link

Photographer

Sean Baker Photo

Posts: 8044

San Antonio, Texas, US

Robb Mann wrote:
Sean,

You've moved past photoshop. Check out this package instead: http://www.ittvis.com/

The lack of a price anywhere on that website tells me all I need to know about whether I'll ever get that purchase past the House-6.  Still, something to consider if school doesn't work out smile.

Apr 24 09 10:03 am Link

Photographer

Art House

Posts: 204

Cincinnati, Ohio, US

Sean, I just wanted to add to the appreciation and many thanks you have already received. I was very intrigued when you first posted this technique, but didn't have the time for experimentation until today. What a fantastic technique!! The control is amazing. I'll definitely be adding this technique into the arsenal. It's posts like these that make this place worth while.

Thanks again.

Aaron

Apr 24 09 10:54 am Link

Photographer

Nicholas Cooper

Posts: 653

Portland, Oregon, US

The instructions posted are not going to be concrete for all images though. You forgot to calculate things in such as contrast across the photo and other details. It's a decent starting point but people are going to have to experiment with the values. It is a good starting point though and I will check it out and experiment with it to see if it's comparable to current methods I use.

The values posted will vary on the image. It's like saying "I have this high powered Mustang with wide 335s in the back" but when you try to put those big ass tires on a Camry, good luck. It may work for one scenario but not another.

Apr 24 09 12:40 pm Link

Photographer

Sean Baker Photo

Posts: 8044

San Antonio, Texas, US

Nicky TwoThumbs wrote:
The instructions posted are not going to be concrete for all images though.

I'm sorry - where did you see me posting pixel values applicable to everyone's images?  Let me know - I need to revise a post if so.

Apr 24 09 12:52 pm Link

Photographer

Nicholas Cooper

Posts: 653

Portland, Oregon, US

Sean Baker wrote:

I'm sorry - where did you see me posting pixel values applicable to everyone's images?  Let me know - I need to revise a post if so.

Hold it sport, don't take it the wrong way. I do recall you said that the mathematical way the program handles things is fucked up to a degree, but then you throw numbers out there like it's going to just WORK for everything with no disclaimer. This is the reason we have people out there creating horrible HDR images as an example because they just move sliders to values that other people list in their own examples or tutorials. Applying an action with predefined values and settings is a starting point but can be a crutch for some people.

And while I do appreciate your posting it as something different to try, I never meant for you to think I was attacking you. There is never any right or wrong way to do it in Photoshop, but there are hundreds of different ways something can be done in Photoshop. Your example is another to add that will work.

Apr 24 09 01:30 pm Link

Photographer

Shutterslam

Posts: 224

Sudbury, Massachusetts, US


That is a matter of taste, and depends on subject content, capture sharpening (if applied), detail in the image, and your output method & size.  I can't give you a single number which will be best for anything you do.  I'll tell you that when I work with an 800x530 image, I tend to use ~.6 or .8.  When I work at full resolution, it's usually scaled from that (so for a 4200px image I would first try 3.6), but I always tune it to the image.  There are other folks here, though, who can speak to output sharpening far better than I - printing is still something which is very new for me.

* Ah, ok that's where I went wrong...I overshot the GB values

This is partly off topic, but I wanted to share a small story with you.

My fiance's father in law passed away just recently.  I was asked to digitally reproduce about 30 photos spanning his life in order to assemble a slideshow intended to play at his memorial...apropos considering I'm never without my camera and laptop.

With your sharpening technique and a toolbox of other correction techniques I was able to restore these photos to a previously unimagineable level of detail and quality.  What would have been barely identifiable photos were saved.  It was a somber enough moment, but when that slideshow was done, there was barely a dry eye in the chapel.

I wish I could share those photos with you, but you'll have to settle for a heartfelt thank you for helping this family say goodbye in a wonderfully extraordinary and unique way.

Cheers.

Lito

Apr 24 09 02:05 pm Link

Photographer

Sean Baker Photo

Posts: 8044

San Antonio, Texas, US

Nicholas Cooper wrote:
I do recall you said that the mathematical way the program handles things is fucked up to a degree, but then you throw numbers out there like it's going to just WORK for everything with no disclaimer.

For clarification, as this seems to be causing some confusion, there are only two numbers in the technique instructions for each editing mode.  Both are universal and explained below.  The actions use these numbers automatically, while allowing you to select the radius at which you seek to separate the image through the normal Gaussian Blur dialog.

   '2' in the scale box (both).  During an 'add' function this averages the two source images.  Works for every image which has followed the directions.  Maybe this can screw up if you're using CS4 (higher GB limit) and blurring past 1/4 or 1/2 of the image dimensions; I doubt this.
   '128' in the offset box (8bit).  It's 50% grey; I don't know how much more innocuous / universal that number can be (within 8 bit; it doesn't work as well in 16bit).
   '0' in the offset box (16bit).  In 16bit mode, the added inverse image (adding a negative; subtracting) requires no offsetting and is left alone.

Shutterslam wrote:
I wish I could share those photos with you, but you'll have to settle for a heartfelt thank you for helping this family say goodbye in a wonderfully extraordinary and unique way.

You are most assuredly welcome.  Thank you for sharing that.

Sean

Apr 24 09 07:09 pm Link