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

Photographer

grahamsz

Posts: 1039

Boulder, Colorado, US

Soliz Photography wrote:
Alright I read into what this is all about, very interesting to say the least.
Now with the Action Script 'Frequency.atn'

The High Frequency layer and Low Frequency layer, which tools do I use on which layer? I started to do some test healing on the High and Low, noticed they both did close to the same thing.

Still going to need to back track to understand, but any other information would be great.

-NTS

Try painting onto the low frequency layer so and get a feel for how you can smooth out tones without removing the pores and skin texture on the high frequency layer.

Then try taking the clone tool on 100% opacity and try cloning an area of good skin over a blemish. Even if the underlying tone of the two areas is different, you can still clone between them because the tone should be on the LF layer.

Generally i use a 5-10px blur on the low frequency layer. I think the scripts preset is a little too low for my taste.

May 07 09 03:42 pm Link

Photographer

Soliz Photography

Posts: 22

Palo Cedro, California, US

grahamsz wrote:

Try painting onto the low frequency layer so and get a feel for how you can smooth out tones without removing the pores and skin texture on the high frequency layer.

Then try taking the clone tool on 100% opacity and try cloning an area of good skin over a blemish. Even if the underlying tone of the two areas is different, you can still clone between them because the tone should be on the LF layer.

Generally i use a 5-10px blur on the low frequency layer. I think the scripts preset is a little too low for my taste.

Thanks, I played some more with it, and yes I bumped the blur to +5.5
Again thanks.

May 07 09 03:58 pm Link

Photographer

Kevin Connery

Posts: 17690

El Segundo, California, US

grahamsz wrote:
Generally i use a 5-10px blur on the low frequency layer. I think the scripts preset is a little too low for my taste.

Resolution-dependent.

The same shot (closeup of a face, for example) at 5 megapixels vs 20 megapixels will benefit from a different radius, as the 20 MPixel one has 4 pixels for every 1 the 5 MPixel file has, and a different radius will be needed to obtain the same effects.

It's also content-dependent. Images from the same camera, lens, and subject may benefit from different values if the distance changes--a full-body shot vs a close-up.

May 07 09 04:42 pm Link

Photographer

Ronald N. Tan

Posts: 2747

Los Angeles, California, US

Friendly bump. smile I like the stuff in here.

May 07 09 04:46 pm Link

Photographer

Photons 2 Pixels Images

Posts: 17007

Berwick, Pennsylvania, US

Kevin Connery wrote:

Resolution-dependent.

The same shot (closeup of a face, for example) at 5 megapixels vs 20 megapixels will benefit from a different radius, as the 20 MPixel one has 4 pixels for every 1 the 5 MPixel file has, and a different radius will be needed to obtain the same effects.

It's also content-dependent. Images from the same camera, lens, and subject may benefit from different values if the distance changes--a full-body shot vs a close-up.

This is what I've noticed also. And why I am contemplating what I mentioned earlier: Drag a rectangular selection around the face of the main subject or something similar in real world size. Use the area of the selection to determine the approximate radius of blur. I'm thinking it would get you real close most times.

May 07 09 05:08 pm Link

Photographer

Ruben Vasquez

Posts: 3115

Puyallup, Washington, US

Kevin Connery wrote:
Resolution-dependent.

The same shot (closeup of a face, for example) at 5 megapixels vs 20 megapixels will benefit from a different radius, as the 20 MPixel one has 4 pixels for every 1 the 5 MPixel file has, and a different radius will be needed to obtain the same effects.

It's also content-dependent. Images from the same camera, lens, and subject may benefit from different values if the distance changes--a full-body shot vs a close-up.

I had a feeling these were some important factors to consider. Can you or any one else here recommend some general guidlines or rule(s) of thumb?

May 07 09 08:33 pm Link

Photographer

Photons 2 Pixels Images

Posts: 17007

Berwick, Pennsylvania, US

Ruben Vasquez wrote:

I had a feeling these were some important factors to consider. Can you or any one else here recommend some general guidlines or rule(s) of thumb?

My rule of thumb right now to get me close is the area of the face in pixels. I'll use my camera for an example: 3888 X 2592 pixels. Let's say the face takes up approximately 1000 X 800 pixels. I'd use a radius of about 12. Suppose I shorten the focal length and take a 3/4 body shot and now the face is only about 400X300 pixels. I'd go with a radius of about 3 or 4.

These numbers are just examples as I haven't actually tried to figure out the best ratio yet. Also, depending on the skin itself, you may want to go higher or lower with the amount of blur.

Basically, just blur until the skin tones mostly even out and most of the imperfections are gone. Then start your healing/cloning.

May 08 09 03:01 am Link

Photographer

Sean Baker Photo

Posts: 8044

Fairfax, Virginia, US

Ruben Vasquez wrote:
I had a feeling these were some important factors to consider. Can you or any one else here recommend some general guidlines or rule(s) of thumb?

You may also find that running the actual HP filter once before you get started can help you find where (ish) the imperfections lie.  Start at a very small radius and keep raising it until you see good detail without the appearance of blemishes - note that radius as it's what you'll use to separate.  Keep going until the blemishes are present but don't get any worse - this will be your end point for the second blur in grahamsz' technique.  Cancel out, separate, and do your second blur.

May 08 09 08:22 am Link

Photographer

Photons 2 Pixels Images

Posts: 17007

Berwick, Pennsylvania, US

If anyone is feeling adventurous, I got something for you to try. It's a variation on the separation method in this thread.

This is set to work on RGB/16 bit only!

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

Load that into your actions. Run the action "Sharpen" in that action set.

The first GB dialog is for the skin tones layer and will be inverse masked. It's set to 25 by default.

The second, third, and fourth GB dialogs are used for your separations. They are set to 2.5, 5.0, and 10.0 respectively by default. You can set any of these to whatever you want, really. Just keep in mind you will be working with 3 layers of separation instead of just one.

After the action completes, you will have your LF layer on which you can do the low level touchup if needed. This layer is blurred to whatever you chose in the last GB dialog. You will have a blank layer above this for any color corrections or anything else you may need. Above this is your blurred skin tones layer. Then you have H1, H2, and H3 each of which are the applied "high frequency" layers for each of the 2nd, 3rd, and 4th GB dialogs. All these layers are placed into a group named "Freq".

Above these, you will notice a copy of each of the H1, H2, and H3 layers placed in a group named "Sharp Masks" with inverted masks.

This is how I worked through this setup:

1. Start with the LF layer and do any cloning, touchup, healing, etc. needed.
2. Work the skin tones layer and blank layer to get the overall tones of the skin the way you want.
3. Make visible HF1 and select that layer. Do any cloning, healing, and touchup you need on this layer to make everything look as nice as possible. You will notice on this layer that most of the skin is blurred with edges sharp.
4. Make HF1 invisible and HF2 visible. Do the same thing on this layer that you did on HF1.
5. Make HF2 invisible and HF3 visible. Repeat same as the previous 2 layers.
6. Make HF1 and HF2 visible. This will look pretty bad, actually. Lower opacity on HF1 and HF2 to tune in the right amount of effect for what you want. I set mine to around 35% each. You will have to tune to taste.
7. Use the copy of HF1, HF2, and HF3 and unmask areas that you want to sharpen up just a bit more such as eyes, lips, hair, etc.

You can turn the groups on and off to see the overall effect of what you've done compared to the original.

If anyone finds this useful or thinks of anything I missed that would make it better, please let me know.

May 09 09 08:15 pm Link

Photographer

AC Fashion Fusion

Posts: 134

Toronto, Ontario, Canada

Thanks for this.  My buddy emailed the discussion to me and I am most grateful. Now I have more elements of PS3 exposed.  Thank you.  I am one of those Hacks someone mentioned earlier.  I hacked my way through grad school and kinda like the sound of my own hacking.
Thank you so much.

May 10 09 05:19 pm Link

Photographer

StudioCMC II

Posts: 487

Bountiful, Utah, US

Robert Randall wrote:

When you've hung out in here long enough, you come to realize there are a large number of people that can't accomplish much in life. They need some form of reinforcement in their lives, and they seem to have chosen the art card for help with that. They can't make art, but they can hide behind the art card. Their arguments use the ploy of deflection to acheive credibility, much the same that OJ did during his murder trial. Everyone knows they are talentless hacks, but when they produce a snap shot of their girlfriend tugging at her soiled panties while sticking her ass out and giving the cell phone that come hither look that reminds you of a deer about to meet it's maker, they argue the picture is their art, and that somhow validates everything, and they are no longer a talentless hack screwing a toothless hag, they am be ARTEESTS!

Of course in this particular case, I'm sure the respondent is a highly talented artist that has a firm belief in the soul of his work. 

Or not.

OMG.. this is TOO real.. Stop it!! LMFAO!

May 10 09 05:28 pm Link

Photographer

grahamsz

Posts: 1039

Boulder, Colorado, US

Photons 2 Pixels Images wrote:
If anyone is feeling adventurous, I got something for you to try. It's a variation on the separation method in this thread.

That looks pretty cool, i'll try to check it out this evening and report back

May 11 09 07:52 am Link

Photographer

Bennett Shoots Fashion

Posts: 98

Philadelphia, Pennsylvania, US

This thread (and Koray's) is great! I love using this new method on skin, much nicer than high pass!

May 12 09 01:11 pm Link

Photographer

Photons 2 Pixels Images

Posts: 17007

Berwick, Pennsylvania, US

grahamsz wrote:

That looks pretty cool, i'll try to check it out this evening and report back

I've been playing with this for awhile and I kinda like it. It gives me a bit more control over the details.

If you can think of anything else this might need, definitely let me know.

May 13 09 07:20 pm Link

Photographer

grahamsz

Posts: 1039

Boulder, Colorado, US

I'll let you know. I've been a bit too busy with work and other shit to get to this sad Doing some more shoots at the weekend so i'm sure i'll try it out on those results.

May 13 09 08:17 pm Link

Photographer

grahamsz

Posts: 1039

Boulder, Colorado, US

Well, honestly, i'm happier using your original action. The multistep process seems to work well, but it's slower and a bit more cumbersome to work with.

I'll keep you posted as I experiment more

May 17 09 12:02 am Link

Photographer

Sean Baker Photo

Posts: 8044

Fairfax, Virginia, US

I have a bit of time again, so I'd like to run through a few uses for these techniques / theories which haven't yet been discussed as well as expound on a few other things I've mentioned.  Whether you find them to of value is a valid question; if nothing else, hopefully the exercise will help you to learn / discover / create something new.

Blended / graded selection of frequency-based elements:
  1.) Use the technique as outlined at the beginning of the thread.
  2.) Above the high frequency layer, create a new empty layer.
  3.) Fill with 50% gray.  Change blend mode to 'Difference'.
  4.) Select all. (Ctrl/Cmd+A)
  5.) Copy merged. (Ctrl/Cmd+Shift+C)
  6.) Go to the channels palette.
  7.) Create a new channel.
  8.) Paste what you just created into this new channel. (Ctrl/Cmd+V)
  9.) Load the selection (Ctrl/Cmd+Click) and play.

  *) Used at lower frequencies, this is a neat alternative to 'Orton'-esque techniques allowing you freedom to manipulate highs and lows at your choice of spatial freq.  At high frequencies you have a strange variation on edge-detection type filters which give you a graded selection of the high-frequency elements in the scene.  For something really neat, try selecting everything in the channel except the black and then invert.  Run a sharpening filter against the resulting selection to even the apparent sharpening at the edges of your fine details.


Selection of intermediate frequencies:
  This could be inferred from some of what has been discussed before, but for the sake of clarity I'll run through it quickly.  I'll assume 16bit editing; refer to the OP for the differences which would allow you to adapt this to 8bit.
  1.) Create two duplicates of the image to be worked on.  We'll refer to these as the upper + lower layers.
  2.) Apply Gaussian Blur to the lower layer at the lower frequency which you want to be a part of your selection.
  3.) Select the upper layer.
  4.) Run the Apply Image command. (Ctrl/Cmd+Shift+A)
    4a.) Choose the lower layer as the source layer.
    4b.) Check the 'Invert' box.
    4c.) Choose 'Add' for the blending mode.
    4d.) Opacity: 100%; Scale: 2; Offset: 0.
    4e.) Click OK.
  5.) Discard / delete / hide the lower layer.
  6.) Select the upper layer.
  7.) Apply Gaussian Blur at the higher frequency which you want to be a part of the selection.
  8.) Refer to the steps above to continue, treating this new layer as if it were the HP data.

Noise-based replacement of removed spatial frequencies in skin.
  This is a simple first foray into writing a PS Script (P2P inspired me smile ) which will generate a largely continuous noise pattern over desired spatial frequencies.  It's far from perfect, but at the default settings (edit it to change them; I tried to make it easy-ish) will create a fairly regular noise pattern within the frequencies I find myself most needing to eliminate from skin.  There are no 'protections' built into this version, so please be careful using this as you could very well crash PS with it - you have been warned.
  A few things to note if you try this:
  1.) When choosing a starting size, you must choose a scaling factor which will round up to at least one more pixel in dimension.  Otherwise you'll never actually scale the image.
  2.) Choosing too large a starting size, too large a scaling factor, or too many iterations will result in an image of gigantic proportions which could crash PS.  The default settings will produce an image ~2800px on a side.
  3.) The noise amount will heavily impact how quickly / whether you bottom / top out any region in the image (think of it as a blown highlight from noise).  You can rerun the script of just use the healing brush to fix this.

On automated skin smoothing, blurring, and edges:
  1.) The techniques discussed throughout this thread are all fantastic.  The world doesn't like talking about such things (can't bill for as many of your hours of hard labor), but these are all automated ways of doing the same thing as pixel-level D&B.  Whether you use them correctly, appropriately mask the result, and exhibit restraint will determine whether you get results which look like a master or a GWC.
  2.) One of the problems with repeat applications of Gaussian Blur on lower frequency data is that the repeat loss of intermediate information contributes to a loss of the strong edges in the image.  That is, if you think about grahamsz original explanation of multiple frequencies contributing to a composite 'edge' output, if we remove the intermediate frequency (as these techniques are doing) we also lose some of our edge contrast.
     2a.) One way of dealing with this is to just avoid applying the techniques to the edges of the skin.  This works wonderfully unless there are skin elements which we want to avoid at the periphery.
     2b.) Another technique would be to sharpen the lower-frequency data in order to 'firm up' the border.  This can best be done with the Smart Sharpen filter, but alas you'll be limited to a maximum 64px radius.  I'm sure there are mathematical ways of calculating the ideal settings for doing this, but at risk of further ruining photography for all of you ( wink ) I'll refrain from discussing them.
     2c.) We might also consider alternative ways of performing our original or secondary blurs.  By using either the Median or Surface Blur filters, we can obliterate the offending spatial data while using a method which is meant to respect strong borders (and does so very well).
     2d.) Finally, we can utilize one of those 13 channels available to us in PS which Mr. Randall and Mr. Connery speak of so often.  If we use the one with the greatest contrast in the skin borders to create a selection in which to originally perform our blurring operations, PS will have no choice but to respect our borders.  Do note that this will work much better for secondary blur operations than it will for primaries.
  3.) I advocate, and wrote the above script, for my belief that when one removes a range of spatial data from an edited image, we risk creating a new problem for ourselves.  When the resultant image is resized, we move from an original state in which we had spatial data appearing continuously across the image.  With a segment of that now gone, resizing can result in destructive interference among the remaining skin elements and the appearance at smaller size of that 'Barbie' skin which most of us are trying to avoid.  Replacement of the spatial frequency range largely prevents this from happening.  [Note that the occurrence of this phenomenon will only occur with certain resizing ratios + ranges of skin elements removed].  To see a discussion on destructive interference & image resizing, please see Mr. van der Wolf's work here.


So, I realize that was a whole lot to digest in having written it so hurriedly I'm certain there are plenty of mistakes in it.  If you see them, please say something so that I might correct myself (there will be edits; so sorry if that bothers you).  And if you have questions, always ask.

May 17 09 10:05 am Link

Photographer

grahamsz

Posts: 1039

Boulder, Colorado, US

Sean Baker wrote:
3.) I advocate, and wrote the above script, for my belief that when one removes a range of spatial data from an edited image, we risk creating a new problem for ourselves.  When the resultant image is resized, we move from an original state in which we had spatial data appearing continuously across the image.  With a segment of that now gone, resizing can result in destructive interference among the remaining skin elements and the appearance at smaller size of that 'Barbie' skin which most of us are trying to avoid.

Personally my vote is for attenuation.We can remove 80% of the offending frequencies, smooth up the skin and yet keep something there for the brain to latch onto as it scales. I'm just not convinced that throwing noise in to fill the space is necessarily the best approach.

Of course this isn't something i do wholesale across the image. I find it works best to selectively remove those middle frequencies on legs, arms and sometimes foreheads. Also was able to use a variant of it yesterday to remove some of the models ribs.

I suppose i should try to find some good comparative images, process them different ways and hold some kind of poll. It'd be interesting to see if certain looks are perceived as better.

May 17 09 10:35 am Link

Photographer

Sean Baker Photo

Posts: 8044

Fairfax, Virginia, US

I certainly don't mean to advocate that any of these techniques is appropriate to every situation, nor that noise-based replacement of removed data is always the most important.  If I've given that impression then I've misspoken and apologize.  Rather, I want to share some thoughts on ways which one might find useful in some situations.

I try to use the shortest route to the best result with skin editing [as far as my miniscule right brain can perceive it to be at least].  Sometimes that's a Surface Blur + noise replacement; sometimes attentuation; sometimes RNT-style D&B.  Noise-based replacement, IMO, is best applied when a large swath of information has been harshly removed from the original image.  Say a factor of 10 (i.e. 5-50px) removed.  For smaller ranges, the likelihood of visible problems is rather low.  Even at that, though, a noise pattern like that created by the script - when embossed and blurred to match the scene (namely, light angle + depth) - can give a reasonable sense of pore detail if the user lacks other sampling sources to copy from.  Doing so on a smart filter makes it adaptable as you go.  Lastly, displacement mapping of the same layer can add to the sense that the detail is really a part of the scene and is a natural part of the skin.

These are only a few thoughts and meant more as discussion pieces than advice or absolute truths.  Only the OP - that the High Pass filter isn't as accurate as it should be - is fact smile.

May 17 09 03:14 pm Link

Photographer

Jerry Bennett

Posts: 2223

Philadelphia, Pennsylvania, US

Sean Baker wrote:
Lastly, displacement mapping of the same layer can add to the sense that the detail is really a part of the scene and is a natural part of the skin.

Hmmmm... I never thought of doing that to a noise layer....

May 17 09 08:16 pm Link

Photographer

Photons 2 Pixels Images

Posts: 17007

Berwick, Pennsylvania, US

Sean Baker wrote:
I have a bit of time again, so I'd like to run through a few uses for these techniques / theories which haven't yet been discussed as well as expound on a few other things I've mentioned.  Whether you find them to of value is a valid question; if nothing else, hopefully the exercise will help you to learn / discover / create something new.

Blended / graded selection of frequency-based elements:
  1.) Use the technique as outlined at the beginning of the thread.
  2.) Above the high frequency layer, create a new empty layer.
  3.) Fill with 50% gray.  Change blend mode to 'Difference'.
  4.) Select all. (Ctrl/Cmd+A)
  5.) Copy merged. (Ctrl/Cmd+Shift+C)
  6.) Go to the channels palette.
  7.) Create a new channel.
  8.) Paste what you just created into this new channel. (Ctrl/Cmd+V)
  9.) Load the selection (Ctrl/Cmd+Click) and play.

  *) Used at lower frequencies, this is a neat alternative to 'Orton'-esque techniques allowing you freedom to manipulate highs and lows at your choice of spatial freq.  At high frequencies you have a strange variation on edge-detection type filters which give you a graded selection of the high-frequency elements in the scene.  For something really neat, try selecting everything in the channel except the black and then invert.  Run a sharpening filter against the resulting selection to even the apparent sharpening at the edges of your fine details.


Selection of intermediate frequencies:
  This could be inferred from some of what has been discussed before, but for the sake of clarity I'll run through it quickly.  I'll assume 16bit editing; refer to the OP for the differences which would allow you to adapt this to 8bit.
  1.) Create two duplicates of the image to be worked on.  We'll refer to these as the upper + lower layers.
  2.) Apply Gaussian Blur to the lower layer at the lower frequency which you want to be a part of your selection.
  3.) Select the upper layer.
  4.) Run the Apply Image command. (Ctrl/Cmd+Shift+A)
    4a.) Choose the lower layer as the source layer.
    4b.) Check the 'Invert' box.
    4c.) Choose 'Add' for the blending mode.
    4d.) Opacity: 100%; Scale: 2; Offset: 0.
    4e.) Click OK.
  5.) Discard / delete / hide the lower layer.
  6.) Select the upper layer.
  7.) Apply Gaussian Blur at the higher frequency which you want to be a part of the selection.
  8.) Refer to the steps above to continue, treating this new layer as if it were the HP data.

Noise-based replacement of removed spatial frequencies in skin.
  This is a simple first foray into writing a PS Script (P2P inspired me smile ) which will generate a largely continuous noise pattern over desired spatial frequencies.  It's far from perfect, but at the default settings (edit it to change them; I tried to make it easy-ish) will create a fairly regular noise pattern within the frequencies I find myself most needing to eliminate from skin.  There are no 'protections' built into this version, so please be careful using this as you could very well crash PS with it - you have been warned.
  A few things to note if you try this:
  1.) When choosing a starting size, you must choose a scaling factor which will round up to at least one more pixel in dimension.  Otherwise you'll never actually scale the image.
  2.) Choosing too large a starting size, too large a scaling factor, or too many iterations will result in an image of gigantic proportions which could crash PS.  The default settings will produce an image ~2800px on a side.
  3.) The noise amount will heavily impact how quickly / whether you bottom / top out any region in the image (think of it as a blown highlight from noise).  You can rerun the script of just use the healing brush to fix this.

On automated skin smoothing, blurring, and edges:
  1.) The techniques discussed throughout this thread are all fantastic.  The world doesn't like talking about such things (can't bill for as many of your hours of hard labor), but these are all automated ways of doing the same thing as pixel-level D&B.  Whether you use them correctly, appropriately mask the result, and exhibit restraint will determine whether you get results which look like a master or a GWC.
  2.) One of the problems with repeat applications of Gaussian Blur on lower frequency data is that the repeat loss of intermediate information contributes to a loss of the strong edges in the image.  That is, if you think about grahamsz original explanation of multiple frequencies contributing to a composite 'edge' output, if we remove the intermediate frequency (as these techniques are doing) we also lose some of our edge contrast.
     2a.) One way of dealing with this is to just avoid applying the techniques to the edges of the skin.  This works wonderfully unless there are skin elements which we want to avoid at the periphery.
     2b.) Another technique would be to sharpen the lower-frequency data in order to 'firm up' the border.  This can best be done with the Smart Sharpen filter, but alas you'll be limited to a maximum 64px radius.  I'm sure there are mathematical ways of calculating the ideal settings for doing this, but at risk of further ruining photography for all of you ( wink ) I'll refrain from discussing them.
     2c.) We might also consider alternative ways of performing our original or secondary blurs.  By using either the Median or Surface Blur filters, we can obliterate the offending spatial data while using a method which is meant to respect strong borders (and does so very well).
     2d.) Finally, we can utilize one of those 13 channels available to us in PS which Mr. Randall and Mr. Connery speak of so often.  If we use the one with the greatest contrast in the skin borders to create a selection in which to originally perform our blurring operations, PS will have no choice but to respect our borders.  Do note that this will work much better for secondary blur operations than it will for primaries.
  3.) I advocate, and wrote the above script, for my belief that when one removes a range of spatial data from an edited image, we risk creating a new problem for ourselves.  When the resultant image is resized, we move from an original state in which we had spatial data appearing continuously across the image.  With a segment of that now gone, resizing can result in destructive interference among the remaining skin elements and the appearance at smaller size of that 'Barbie' skin which most of us are trying to avoid.  Replacement of the spatial frequency range largely prevents this from happening.  [Note that the occurrence of this phenomenon will only occur with certain resizing ratios + ranges of skin elements removed].  To see a discussion on destructive interference & image resizing, please see Mr. van der Wolf's work here.


So, I realize that was a whole lot to digest in having written it so hurriedly I'm certain there are plenty of mistakes in it.  If you see them, please say something so that I might correct myself (there will be edits; so sorry if that bothers you).  And if you have questions, always ask.

Very interesting. I'm going to try this out more later when I have time to sit down and can let it sink in.

Should this thread be moved to the new "Digital Art and Retouching" category? I'd like to see it there as it's very useful.

May 23 09 06:34 am Link

Photographer

Robert Randall

Posts: 13860

Chicago, Illinois, US

I just re-read the thread to gain a better understanding of the skin retouching potential. One thing that stood out is the ability to enhance the impact of the high frequency (Detail) layer with a curve. You can stand on this with all your weight and just watch the detail explode. It becomes particularly effective if you hit the Detail layer with a slight amount of GB to soften the effects, there by keeping it looking realistic (at least to me). I'm working on files that are 5616 X 3744 from a Mark III. I used .4 GB to kill the artifacting I noticed from a completely clamped curve.

Phenomenal stuff!

May 30 09 12:23 pm Link

Photographer

Photons 2 Pixels Images

Posts: 17007

Berwick, Pennsylvania, US

Robert Randall wrote:
I just re-read the thread to gain a better understanding of the skin retouching potential. One thing that stood out is the ability to enhance the impact of the high frequency (Detail) layer with a curve. You can stand on this with all your weight and just watch the detail explode. It becomes particularly effective if you hit the Detail layer with a slight amount of GB to soften the effects, there by keeping it looking realistic (at least to me). I'm working on files that are 5616 X 3744 from a Mark III. I used .4 GB to kill the artifacting I noticed from a completely clamped curve.

Phenomenal stuff!

I agree completely. Phenomenal stuff here. I've been making a copy of the High Frequency layer, inverted mask, then paint back in over the eyes to bring out a bit more detail.

Maybe not the greatest example here, but using this method (I have all the layers set up using an action so I don't have to create the layers) I can go from this....

http://www.nunuvyer.biz/miscimages/Eye.jpg

to this...

http://www.nunuvyer.biz/miscimages/EyeC.jpg

in about 10 seconds.

May 30 09 05:43 pm Link

Photographer

DIGITAL SNIPER

Posts: 200

Rochester, New York, US

rock on great atn

May 30 09 08:02 pm Link

Photographer

DIGITAL SNIPER

Posts: 200

Rochester, New York, US

I used the new sharpen !

http://photos.modelmayhem.com/photos/090530/20/4a21fd797ce7d_m.jpg

May 30 09 08:56 pm Link

Photographer

GavinJPhoto

Posts: 93

Spokane, Washington, US

Paul Dempsey wrote:
sounds to me like you're taking the "art" out of photography and turning it into a math equation....ther'e s abook you should check out ..."Drawing on the Right Side of the Brain"  It's designed to help people get away from the over analytical thinking and rely more on the creative side (the right side ) of your brain...plus, it's just way more fun to look at photography as an art form rather than a math problem....

....if an image does not contain the elements of a good photograph:ie, composition, color &l ight etc., no amount of sharpening (of any type) will make it a good photograph
I imagine if MM's best photographers were to critique all the photos on MM that "used the wrong sharpening tool" would be one of the least frequent comments or recommendations for improvement.
So, for many photogs, sharpening techniques are great, but they are not a shortcut to great photos.  I have seen too many photogs get hung up on the technical (or math) side of photography, while, at the same time ignoring the artistic side.
Obviously Sean has hit on something that has much value to many of the photogs here.   I congratulate him for sharing his technique in this forum.
My comments are in no way an attempt to disparage his good work and generosity.

Paul

The way your photos are created and processed is mathematics my friend. lol

Jun 12 09 05:47 pm Link

Photographer

Robert Randall

Posts: 13860

Chicago, Illinois, US

I thought I would share the success I feel I've had with Sean's deconvolution formula. I just posted a new web site, one of those high-def flash jobs that show every little detail and flaw. I used Sean's process on every single one of the files on the site, and I couldn't be happier. However, the reason I writing this has more to do with a side effect of the process that makes me the happiest of all.

On the web site is a menu item titled "Branding Campaigns", and a sub menu of that item is titled "Mid West Bank". Every shot in that folder was made at ISO 1200, and I can't believe how crisp and beautiful they look because of the process.

http://www.robert-randall.com/content/

So thank you Sean and everyone else that experimented and shared this process, this was a big one!

Jun 17 09 05:18 pm Link

Photographer

BornArts

Posts: 305

Fresno, California, US

DigitalEdge Photography wrote:

The way your photos are created and processed is mathematics my friend. lol

Woooosh!  Math has boundaries, art is limitless, and while the universe revolves around mathmetics, it is the art that made you pick up a camera.

Jun 17 09 05:41 pm Link

Digital Artist

Koray

Posts: 6716

Ankara, Ankara, Turkey

Robert Randall wrote:
I thought I would share the success I feel I've had with Sean's deconvolution formula. I just posted a new web site, one of those high-def flash jobs that show every little detail and flaw. I used Sean's process on every single one of the files on the site, and I couldn't be happier. However, the reason I writing this has more to do with a side effect of the process that makes me the happiest of all.

On the web site is a menu item titled "Branding Campaigns", and a sub menu of that item is titled "Mid West Bank". Every shot in that folder was made at ISO 1200, and I can't believe how crisp and beautiful they look because of the process.

http://www.robert-randall.com/content/

So thank you Sean and everyone else that experimented and shared this process, this was a big one!

if images like those doesnt get one hired I dont know what else would.
good stuff man.

Jun 17 09 06:15 pm Link

Photographer

Robert Randall

Posts: 13860

Chicago, Illinois, US

Belk Media Group wrote:

Woooosh!  Math has boundaries, art is limitless, and while the universe revolves around mathmetics, it is the art that made you pick up a camera.

Math is as boundless as your imagination, the universe alternately revolves around the tip of a pin and a place as incomprehensible as the thought you just posted, and I imagine it is at a minimum, curiosity, that invites you to pick up anything.

However, if it is math that created my photos, I would love to meet her.

Jun 18 09 08:58 am Link

Photographer

Sean Baker Photo

Posts: 8044

Fairfax, Virginia, US

Robert Randall wrote:
On the web site is a menu item titled "Branding Campaigns", and a sub menu of that item is titled "Mid West Bank". Every shot in that folder was made at ISO 1200, and I can't believe how crisp and beautiful they look because of the process.

http://www.robert-randall.com/content/

Thank you Bob for sharing this, and in doing so for reminding me that I'm not pushing myself nearly hard enough. 

And in that vein - has anyone bothered combining what we've talked about for sharpening here (curves on the 'HP' layer) with the sharpening technique I outlined in Kevin's thread?  Remember that, thanks to the accuracy of the technique, you aren't limited to just one spatial frequency.

Jun 18 09 09:29 am Link

Photographer

JeF Briguet

Posts: 119

Chamoson, Valais, Switzerland

Sean Baker wrote:
And in that vein - has anyone bothered combining what we've talked about for sharpening here (curves on the 'HP' layer) with the sharpening technique I outlined in Kevin's thread?  Remember that, thanks to the accuracy of the technique, you aren't limited to just one spatial frequency.

Sounds interesting!

Anyway i've to thank you again. I've been mixing this with my usual workflow a lot lately (it open endless possibilities) and i've to admit you made me save a lot of time (and money).

Like Bob i've published some of my latest work using the "Sean-frequency-split-method" for sharpening or/and correction/pushing contrast/whatever and the results are just perfect.

Like i said previously (deep inside this thread), you opened Pandora's box with this.

smile

Jun 18 09 12:12 pm Link

Photographer

JeF Briguet

Posts: 119

Chamoson, Valais, Switzerland

Robert Randall wrote:
On the web site is a menu item titled "Branding Campaigns", and a sub menu of that item is titled "Mid West Bank". Every shot in that folder was made at ISO 1200, and I can't believe how crisp and beautiful they look because of the process.

http://www.robert-randall.com/content/

I know you are getting a lot of comments on your post-processing workflow (it's justified).

But i wanted to let you know that i for one really like your "eye" on this set. Great framing/compositions. It's not very common in our digital days.

Jun 18 09 12:23 pm Link

Photographer

Justin Foto

Posts: 3622

Alberschwende, Vorarlberg, Austria

So....

I read this thread when it first came out. I kind of figured, blah - so what and ignored it for a while. I didn't think it would give me much better results than I was already getting.

Anyway, I was a bit bored just now, so I thought I'd give it a try.

HOLY SHIT this is good! Definitely part of my work flow from now on.

Jun 18 09 01:02 pm Link

Photographer

Sean Baker Photo

Posts: 8044

Fairfax, Virginia, US

Continuation of a conversation from here.

grahamsz wrote:
I have a gut feeling that I can mitigate that risk by choosing co-prime GB radii. 5 and 15 was probably a horrible choice for an example because there are scaling factors that will smoothly combine them. If you were to choose 5 and 14 then there's no way to scale the result such that there will be whole pixels that match both radii.

That's probably a horrible explanation, but I'm still trying to wrap my head around it.

Actually it makes perfect sense - more intuitively than in a manner I can mathematically prove, but I've been using primes to scale the noise generator for much the same reason.

I'll offer that I've found a similar interplay even without the base numbers directly scaling - it's not as bad, but I started to spot it in my own after looking at enough of them.

grahamsz wrote:
Hmm not sure, i'm scrambling to get cds burned for models before doing further shoots with them. I'll check in this evening and try again

Sounds good... it's sooo probable that I did something wrong in writing that thing, but if you've got a minute I'd like to know where I went astray.

Jun 21 09 10:37 am Link

Photographer

Fun City Photo

Posts: 1552

Stuttgart, Baden-Württemberg, Germany

Sean Baker wrote:
Noise-based replacement of removed spatial frequencies in skin.
  This is a simple first foray into writing a PS Script (P2P inspired me smile ) which will generate a largely continuous noise pattern over desired spatial frequencies.

This script produces an Error 1220 on line 34 docRef.resizeImage (growBy, growBy)

Jun 21 09 11:13 am Link

Photographer

Sean Baker Photo

Posts: 8044

Fairfax, Virginia, US

Fun City Photo wrote:
This script produces an Error 1220 on line 34 docRef.resizeImage (growBy, growBy)

What OS and PS version are you using?  Thanks...

Jun 21 09 11:30 am Link

Photographer

Fun City Photo

Posts: 1552

Stuttgart, Baden-Württemberg, Germany

Vista home,PSCS

Jun 21 09 11:44 am Link

Photographer

Sean Baker Photo

Posts: 8044

Fairfax, Virginia, US

Try this version if you would:

http://www.twicebakedphotography.com/do … eOldCS.jsx

In theory I've found the culprit, but without an older version can't check myself.

Jun 21 09 11:51 am Link

Photographer

Fun City Photo

Posts: 1552

Stuttgart, Baden-Württemberg, Germany

Tried it but it changes from Normal to Soft Light about 20 times and and lets the Noise Layer open.
Will use the scripts Monday in the Studio with CS3 and CS4
Thank you and will tell you about.
BTW I use quite a few very special scripts with little or no use for other photogs.
Like: NEXT IMAGE a script which opens the images on a CF card which are new and have not been opened before, or analize if Portrait or Landscape and resize acordingly. etc.

Jun 21 09 12:13 pm Link