Koray wrote:

I like the sea myself

I think this could use a different set of parameters the more I look at it but nothing serious anyway...

Yes. The sea looks nice too. My attention went right to the rocks and how it brings out every crack.

I retouched an image using the GB filter with a cloudy sky in the background during a sunset. Most of it came out nice, but I wasn't too happy with the clouds/sky. They looked pixelated when I downsized the image no matter what I did short of blurring that area. And me being the "unartist" I couldn't get it to look right to me. The rest of the image I loved. Same here. Just that one small area of the clouds but that's nothing a little masking can't fix, right? That's what I had to do....go back to the beginning and redo the whole image and take care of that area with a mask.

Lumigraphics wrote:
I've been reading Dan Margulis's book and then reading this thread- its interesting that he doesn't have anything along these lines. Four chapters on sharpening, mostly recommending ways to use USM (he dismisses hi-pass sharpening in a sidebar.)

I'm still playing around with all of this but I've already used it for retouching and sharpening.

HP & USM aren't terribly different with the caveat that USM has the advantage of a Threshold whereas HP has the advantage of direct control of highlight / shadow increment and an (IMO) better capacity for being masked.  In all case I advocate inclusion of a deconvolution (smart sharpen) step, but sharpness is very much a matter of personal preference.

...

On another note, I'm looking at using FilterMeister to implement a few things, but am getting the quick sense that it only supports 8bit files.  I know that'd make Bob happy to knock me down a bit ( ), but does anyone know of a free(ish?) filter creation utility which allows implementation of your own code and supports 16 & 32 bit modes?  Edit: And no, I don't have time to learn to use the full SDK, unless someone can both get me approval for access and provide a full wrapper for 16/32bit operations.  Thanks.

Lumigraphics wrote:
I've been reading Dan Margulis's book and then reading this thread- its interesting that he doesn't have anything along these lines. Four chapters on sharpening, mostly recommending ways to use USM (he dismisses hi-pass sharpening in a sidebar.)

I'm still playing around with all of this but I've already used it for retouching and sharpening.

Most people fail to realize that Dan comes from a meat and potatoes color background. His intent is to give the reader/student the ammo necessary to make an image presentable in really crappy print conditions. He is often assailed for the lousy pictures he uses to teach his methods, but those are the kinds of pictures he was faced with on a daily basis in his job as a prepress tech.

It comes as no surprise that he would eschew something like an HP sharpen when all that he ever needed was supplied in a USM filter. It makes even more sense when you only sharpen the black printer.

Sean Baker wrote:
Bottom Line Up Front: High Pass is an inaccurate spatial-frequency based separation technique.  The steps below and the actions provided effect a much more accurate result than can be accomplished otherwise.

Background: Ultimately, this is a spinoff of Mr. Connery's thread on deconvolution as an image sharpening technique here, wherein I set forth my method for finer control of such while both remaining within Photoshop and not spending additional $$ on plugins.  Key to its success is the separation of high and low spatial frequency image data and the fine-tuning of each.

Separation of spatial frequency data has a number of applications in image editing.  Whether for the oft-suggested "High Pass Sharpening", to recover detail lost in an OOF / moving image, or to enhance local contrast throughout the image, the accurate separation of frequency data is relevant for anyone who is a stickler for image quality.

Findings / Technique: In my own experimentation, I've found that HP gives differences as high as 2670/32k per pixel when separating high and low frequency information.  The proposed alternative technique for working with 8bit image data is as follows:

  1.) Start with two copies of the image to be separated.
  2.) Working on the bottom copy, run the gaussian blur filter at the intended pixel frequency (same as you would input into the HP filter).
  3.) Selecting the top copy, choose Apply Image from the Image menu at top (Shift+Ctrl+A on PC; Cmd+Shift+A for Mac).
  4.) In the Layer dialog, select the bottom layer which you blurred in step 2.
  5.) In the Blending dialog, choose Subtract.
  6.) Enter '2' into the Scale box, and '128' for the Offset.
  7.) Preserve Transparency, Mask, and Invert should not be checked.
  8.) Choose OK.
  9.) Your top layer will now look much as a HP result, albeit a bit flatter.  Set the Blend Mode to Linear Light.  Opacity should remain at 100%.

Working in 8bit mode, this will give you an accurate frequency separation to about 115/32k.

Working in 16bit mode, however, we encounter a problem with the offset factor applied in Step 6 above (50% grey in a 16bit system cannot be expressed by an 8bit number).  This can be overcome using the alternate technique which follows:

  1.) Start with two copies of the image to be separated.
  2.) Working on the bottom copy, run the gaussian blur filter at the intended pixel frequency (same as you would input into the HP filter).
  3.) Selecting the top copy, choose Apply Image from the Image menu at top (Shift+Ctrl+A on PC; Cmd+Shift+A for Mac).
  4.) In the Layer dialog, select the bottom layer which you blurred in step 2.
  5.) In the Blending dialog, choose Add.
  6.) Check the Invert box.
  6.) Enter '2' into the Scale box, and '0' for the Offset.
  7.) Preserve Transparency and Mask should not be checked.
  8.) Choose OK.
  9.) Your top layer will now look much as a HP result, albeit a bit flatter.  Set the Blend Mode to Linear Light.  Opacity should remain at 100%.

Conclusion: Accuracy for this technique is fantastic, with a maximum difference from the original of 1/32767.  Unfortunately, this technique is not valid while working in 8bit mode.  I can't explain the phenomenon, but the shadows and midtones get mucked up by about 1-2/256.  Perhaps those with a bit more insight into PS's inner workings can offer thoughts on why this is.

Afterword: Walking through all these steps can get a bit tiring, and I've created an action which I'll link below to run you through them.  It works in a separate window utilizing the 16bit technique above to retain an accurate separation, but, as I'm not an Actions Wizard, it will not move the data back into your original document for you.  If someone more adept than I would like to add that in I will gladly update this post; otherwise it remains to the user to bring both or either layer into your document for use.

What will you do with it?
  - "High Pass Sharpen"?
  - Create a high-spatial-frequency mask?
  - Create a low-spatial-frequency mask?
  - Something else?  Post your results and findings here.  It's time to up the ante on sharing what we find.

Action

Please Note: As outlined above, this technique is designed to end with the image on screen looking exactly the same as it did when you started.  The difference is that your image data is now separated onto two separate layers (separated by size of detail - spatial frequency), allowing you to edit them independently.  If you want to apply "high pass sharpening" with this technique, simply disable the low frequency ('blur') layer and it will appear immediately.

Sean, can I make a suggestion? Would you consider editing your OP and adding in all the links to all the downloads we've made available here? I'm thinking some people may be reading the OP and not getting the rest of what is available to go along with it.

Photons 2 Pixels Images wrote:
Sean, can I make a suggestion? Would you consider editing your OP and adding in all the links to all the downloads we've made available here? I'm thinking some people may be reading the OP and not getting the rest of what is available to go along with it.

Good idea.  I'm working on an image ATM which I want to try to get done tonight, but when I do I'll come back and straighten this out.  Good call.

Sean Baker wrote:
Good idea.  I'm working on an image ATM which I want to try to get done tonight, but when I do I'll come back and straighten this out.  Good call.

No worries, take your time.

I just don't want anyone missing out on anything we've put out here on the off chance they don't read through the whole thread post by post.

I'll go through and compile a list of links to all the posts that contain downloads. I'm not sure if you'd want to link to the post so they get an explanation of what it is along with it, or just link to the downloads themselves and put your own explanation in the OP.

I'll edit this post with the list in just a bit...

Your original action set is in the OP...so no need to relink that one.

My first script/action combination:
https://www.modelmayhem.com/po.php?thre … ost9701584

Second action set that doesn't flatten the image first: (no script included...uses same script as the first one)
https://www.modelmayhem.com/po.php?thre … ost9731906

Syd's Graphic Equalizer:
https://www.modelmayhem.com/po.php?thre … ost9773104

Another multistage separation action set...16 bit only.
https://www.modelmayhem.com/po.php?thre … ost9794067

Your noise generation script:
https://www.modelmayhem.com/po.php?thre … ost9864297

My next step for multistage separation setting up a masked group with HF copies to selectively sharpen.
https://www.modelmayhem.com/po.php?thre … st10223528

And that same one zipped....
https://www.modelmayhem.com/po.php?thre … st10228029

Your noise generations script with the added UI dialog interface:
https://www.modelmayhem.com/po.php?thre … st10360771

I think that's it....don't think I've missed any.

Can you also remind them that the first action set I have linked here has the script packed with it so if they work in anything other than RGB/16 they will want to download that one and use the script to activate whichever action they are using. They will either have to rename the action or edit the script to match.

Thanks, Sean

Edit: Or just quote this post in your OP. Might be easier.

Sean Baker wrote:
Done.  Please let me know if I've missed anything or attributed incorrectly.

WOW! You went a lot more in depth with it than I did.

Thanks!

Photons 2 Pixels Images wrote:
WOW! You went a lot more in depth with it than I did.

Thanks!

I'll be seeing brackets for a week, but it was for sake of a brilliant idea.

Sean Baker wrote:

I'll be seeing brackets for a week, but it was for sake of a brilliant idea.

Just have some more coffee.....you'll be OK.

Speaking of, if I ever get down your way, I'm taking you for a nice cup of coffee and a nice long discussion on some of this stuff. It'll probably end up being a few pots of coffee, though.

Sean Baker wrote:
Were you editing in 8-bit by chance?  I've not been keeping close tabs on P2P's actions and at first glance through this one that's the first thing which comes to mind as an 'easy' reason.

Photons 2 Pixels Images wrote:
That's a good possibility. I only included the script with the first download link. If someone grabbed the one after that, there is only the actions set for 16 bit mode.

If that's what happened, Smedley, just go back to page 5 here and get the original download and you can use that script with the newer actions I've posted. You may just have to rename the actions...I can't remember.

That was it. (editing in 8-bit)  I only noticed it on hi-key images with a white or light solid background. I just tried the same thing in 16-bit and there is no problemo.

For the most part, either action works fine in 8-bit or 16-bit (for me). Actually, they work more than fine, and exceed my expectations most of the time. 

Julian Marsalis wrote:
Ok so where is a video tutorial on this one

I'd be happy to if I knew of a free screen cap program and video editor which produces quality results?  Or better yet, do you know someone at Adobe?

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

Mathematics is a pure art form.

Smedley Whiplash wrote:
I found myself getting lost again a few pages back. Anyone want to clarify GB vs SB for me? I'm using the 2nd Frequency action, and all I see is GB being used. Does another action set use SB ?

Clarification as to how they work, what they're used for, which actions use them, all the above, or something else?

Sean Baker wrote:
I'd be happy to if I knew of a free screen cap program and video editor which produces quality results?  Or better yet, do you know someone at Adobe?

Try Jing @ http://www.jingproject.com/ it's free and pretty sweet for making free videos and has a hosting site for them....

Photons 2 Pixels Images wrote:
Sean, I started looking into doing this for GIMP to try to give some others who don't have access to Photoshop a chance to play with this too. Unfortunately, there is no "Apply Image" in GIMP or anything I can find that is close except for a Convolute filter. If you have any idea how that works and can give me a starting point, I'll see what I can do with it. It seems to get me close, but for the lack of a Linear Light blend mode not close enough. I'll keep looking for some way to do this.

Convolution is a way of degenerating an image - usually blurring (remember Kevin's deconvolution sharpening thread?)... but I'm not aware of any way to use two source images to do that.  In the end you'd just be running a HP and may as well use the built in function.

... thinking as I go...

Actually, have you tested to see if GIMP has the PS issues?  Technically, everything which started this should've been avoided by Adobe years ago in the writing of the filter as it's the artifacts that our technique here avoids.  It's possible that the GIMP dev team never made the same errors / took the same shortcuts as the big guys.

If you have a minute, try to reconstruct an image from a GB and a HP version of the same and see if there are any differences.  I'll try to download a copy and do the same, but you may be able to finish before I even get started.

Sean Baker wrote:
Clarification as to how they work, what they're used for, which actions use them, all the above, or something else?

No... I just haven't seen surface blur used, or I am not identifying it properly?  Someone made the distinction between surface blur and gaussian blur a few pages back, so I just wanted to know the difference as it relates to these actions.

It was Koray, just before he posted the comparison's with NIK's sharpener.
https://www.modelmayhem.com/po.php?thre … st10377649

Sean Baker wrote:
Convolution is a way of degenerating an image - usually blurring (remember Kevin's deconvolution sharpening thread?)... but I'm not aware of any way to use two source images to do that.  In the end you'd just be running a HP and may as well use the built in function.

... thinking as I go...

Actually, have you tested to see if GIMP has the PS issues?  Technically, everything which started this should've been avoided by Adobe years ago in the writing of the filter as it's the artifacts that our technique here avoids.  It's possible that the GIMP dev team never made the same errors / took the same shortcuts as the big guys.

If you have a minute, try to reconstruct an image from a GB and a HP version of the same and see if there are any differences.  I'll try to download a copy and do the same, but you may be able to finish before I even get started.

OK, I'll try.

The convolution is actually rather interesting. They give you a matrix of cells where you enter values to use to figure out the end result.

There is a Difference Of Gaussian filter that allows for 2 radii to be input and it outputs the difference between the two. I think this might be a good starting point.

Edit: I just tried the DOG filter. It actually creates 2 new layers, sets the blur on both, then takes a difference of the 2 and outputs them to another layer, deleting the source layers when finished.

Photons 2 Pixels Images wrote:
OK, I'll try.

The convolution is actually rather interesting. They give you a matrix of cells where you enter values to use to figure out the end result.

There is a Difference Of Gaussian filter that allows for 2 radii to be input and it outputs the difference between the two. I think this might be a good starting point.

Edit: I just tried the DOG filter. It actually creates 2 new layers, sets the blur on both, then takes a difference of the 2 and outputs them to another layer, deleting the source layers when finished.

If it's diff'ing the layers, that's what we need to create the 'HP' layer... any idea what command it's using to do that?  I can't believe how much about GIMP I've forgotten...

Smedley Whiplash wrote:
No... I just haven't seen surface blur used, or I am not identifying it properly?  Someone made the distinction between surface blur and gaussian blur a few pages back, so I just wanted to know the difference as it relates to these actions.

It was Koray, just before he posted the comparison's with NIK's sharpener.
https://www.modelmayhem.com/po.php?thre … st10377649

Gotcha.  To my knowledge, none of the actions currently in the thread are using SB, though I've not looked step by step through each.  It is fantastic (IMO) for use when using the principles of the thread for skin smoothing applied to the low frequency layer, or even as the initial means of separation (a variation on what Koray is talking about doing; just reversed ).  Does that make any more sense?

Sean Baker wrote:

If it's diff'ing the layers, that's what we need to create the 'HP' layer... any idea what command it's using to do that?  I can't believe how much about GIMP I've forgotten...

It does come out looking similar to the HP layer and when I place it on a normal GB layer it comes up close depending on what blend mode I use. Just not quite close enough. Again, no Linear Light blend mode to try with it. I'm thinking that may be the key. I am trying different values for both the LF GB radius and the HF radii, but that doesn't seem to be the problem. It seems to be a problem of blending to the LF layer.

I'm going to see if I can dig up some documentation on it and see what I can figure out as far as the actual command it's using to diff the radii.

Sean Baker wrote:

Gotcha.  To my knowledge, none of the actions currently in the thread are using SB, though I've not looked step by step through each.  It is fantastic (IMO) for use when using the principles of the thread for skin smoothing applied to the low frequency layer, or even as the initial means of separation (a variation on what Koray is talking about doing; just reversed ).  Does that make any more sense?

I wonder if 'surface blur' is a term used in the NIK sharpener software that is the same as gaussian blur, or if it's just Koray's word for the same thing?

i.e. - (NIK has a plugin called Lens Blur that supposedly renders blur more like a lens, instead of a gaussian style blur.)

Sean Baker wrote:
'Grain Extract' and 'Grain Merge' look like they may be the magic blends.  Still doesn't produce a 100% reconstruction, but we're dealing with an 8bit workflow anyway, so I'm not surprised by some small amount of error.  Especially when I can't even find an actual HP command .

I know. I've been looking. I think if we can figure out that Convolute filter, it may give us something usable toward this also. It's basically an empty filter that you populate to tell it how to manipulate each pixel based on surrounding pixels and your parameters.

It looks fairly complex, yet very versatile. The GIMP website gives examples of a sharpen, a blur, edge detect, and some others. This would be more your thing than mine, though.

http://docs.gimp.org/2.6/en/plug-in-convmatrix.html

Sean Baker wrote:
PS has the same but calls it the Custom filter - SE has some nice sharpen implementations for iton his website for download.

WOW! Show's how little I know PS.

I still have trouble with blend modes, though, so it's no wonder.

Robert Randall wrote:

Robert Randall wrote:
Cool, thanks!

I always use my own actions, so I've never had any reason to use the little stop check box thing. No matter how much time I spend with the program, I am always learning new things like that.

Now i get to see how you got sharp without getting sharp!

I don't know that I couldn't do this image with Sean's latest default, but I really like the way you set up your action. I stopped it at the curve and cranked on it a bit more than you did, and then I made up some difference masks from among the material your action supplied. I was able to make a BW file that emulates pretty well the results I would get from using chop masks on a pin registered Fotar enlarger.

I can't tell you guys enough how much this thread has altered my workflow for the better.


http://www.robert-randall.com/Photon2pixels/


Edit for the holy rollers... there are some issues around the glasses and left shoulder of this handsome guy's portrait. Give me a break because I was skittering through it in a hurry, and haven't developed a good work around for that particular problem yet.

Thanks for the example! It always helps to see what others are doing, especially when they know more than I do!