Question to "High pass sucks" technique

[santoro80]

I'm just trying this technique (http://www.modelmayhem.com/po.php?thread_id=439098) and I have two questions to the step 2:

"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)."

1) How do I know what pixel frequency (or radius?) should I choose? I have a 10 Mpx camera.
2) What do they mean by "high pass sucks" or "high pass filter"? This quick de-grunge technique? http://retouchpro.com/tutorials/?m=show&id=213

[mistermonday]

Santoro80,

1) How do I know what pixel frequency (or radius?) should I choose? I have a 10 Mpx camera.

The radius you chooses has nothing to do with the size of your camera output but the content of each image and each image will be different. There are two ways to select the radius. One would be to perform a Gaussian Blur, adjusting the radius slider until all of the detail / fine edges that you want to protect or heal becomes blurred away. The other way is to perform a High Pass Filter adjusting the radius until the detail you want to preserve / heal is revealed. but then click cancel after you determine the radius. The 1st GB method is usually easier to accomplish visually.

2) What do they mean by "high pass sucks" or "high pass filter"? This quick de-grunge technique? http://retouchpro.com/tutorials/?m=show&id=213

The Filter>Other>High Pass has been in PS for 20 years. It was never intended to be precise and because of the math algorithms and other reasons, it is not accurate. So when you create a split frequency set of layers using the conventional HP Filter, you loose quality, visually noticeable particularly in the highlights and shadows.
The method used in that ModelMayhem thread is an alternate way of creating a High Pass layer which is very precise. So if you were just interested in creating only an accurate HP layer, you could create the split layer set and just delete the GB layer. If you compare both methods on a image that has a wide tonal range you will be able to see the difference.

As for the DeGrunge technique, no it is not the same. However the two have some things in common. When you split a layer into spacial frequencies what you effectively have is a formula that says: GB+HP = Original. This will be true at any radius and indeed if you toggle on and off the background layer below the split layers you will see they are visually identical.
For the DeGrunge technique you are actually creating a third component and the following formula: GB + HP + Degrunge = Original. In this case you create a layer where you are protecting all of the fine edges smaller than some radius 1 (HP) and blurring all edges greater than radius 2 (GB) and everything in between represents the grunge. The point to take away from this is that when making a Degrunge layer, you can use the Frequency Split technique to create the HP layer that you would use in the DeGrunge technique. You just make the split then delete the GB layer. Its a few extra steps which you may or may not find worth the extra effort.
Regards, Murray

[santoro80]

Thanks for explanation. It was very helpful!

I just want to use this technique for skin retouching.

[mistermonday]

It works well if you choose the radii carefully and try not do overdo it. If used properly it will save you considerable time instaed of dodging and burning and will not look blurred.
Regards, Murray

[KTG]

I have been experimenting using Inverted High Pass with Apply image. One workflow I am using is making three layers: One with the radius at 3, one with 6 or 9 and one with 18 and selectively applying them to areas of the skin. I read where Natalia uses much larger radii, but I cannot get any effect at a very high radius. Murray is right. If done properly one can reduce greatly the amount of d&b. Which I do after the high pass technique. Any thoughts?

k

[Godmother]

Originally Posted by KTG

I have been experimenting using Inverted High Pass with Apply image. One workflow I am using is making three layers: One with the radius at 3, one with 6 or 9 and one with 18 and selectively applying them to areas of the skin. I read where Natalia uses much larger radii, but I cannot get any effect at a very high radius. Murray is right. If done properly one can reduce greatly the amount of d&b. Which I do after the high pass technique. Any thoughts?

k

Always depends on the quality of the original image. If you're working on a 2000 px wide image a larger than 30 won't work. Most of my image are 8000px wide tho so you can understand :P

[KTG]

Ahhhhh, yes. Thank you GM.

k

[Setcamper]

Trying to wrap my head around what the ModelMayhem thread is actually doing. After completing the 8-bit steps I'm left with 3 layers; my base background image, a layer that's blurred significantly and then a highpass-like layer created using the apply image steps. When all three are turned on I have my original image. What am I supposed to edit or mask to even tones but keep the detail.

I understand how the de-grunge tutorial works because I'm left with a single layer on top of my original that both even tones and holds detail in the areas I un-mask.

[ShadowLight]

depends where you have split the frequencies... but in short you can:

1) pain on the blurred layer to relax shadows/correct colours
2) use heal-brush with settings (current layer only) on the "highpass" layer

[jonjon925]

The blurred layer is the one with the low frequencies of the image and there you can play with the smudge tool to even out things.
The highpass-like layer is the one with the high frequencies of the image and contains texture (facial hair, pores, pimples, hair texture etc) and this is where you use the stamp tool or the healing brush.
These layers are not supposed to be masked out or something like that. You choose the layer you want and you edit it.

[Flashtones]

One use of the split is to degrunge, rather than the inverted highpass, because you get much less edge blurring.

Something like this:

http://www.model-citizens.com/Tutori...paration-1.wmv

[BagLady]

Hi Setcamper,

The low frequency layer (blurred layer) contains color blobs, shadows and highlights.
The high frequency layer contains the detail (no color) detracted from the blurred layer.

These two layers put together make up the original image.... So why split the image in two?

So that you can adjust the details without messing up the underlying color/shadows/highlights and adjust the "color blobs" without ruining the details. Here's a quick explanation on how to do it:

Put these two layers in a group above your original image (so that you can easily see the comparison between this group and your background layer).

Examine the blurred layer first... look for the unwanted blotches/color and correct with whatever method you prefer (healing, cloning, painting, smudging and even a bit of blurring here and there).

Then take a look at the high frequency layer (with all layers visible)... correct unwanted details by healing or cloning.

I hope this helps!

[mistermonday]

Originally Posted by BagLady

Hi Setcamper,

The low frequency layer (blurred layer) contains color blobs, shadows and highlights.
The high frequency layer contains the detail (no color) detracted from the blurred layer.

Just one clarification to what Baglady has written. The high frequency layer actually does contain color (unless you go through a secondary process to move it all to the LF layer). If you wish to see the color and where it resides, just do a Ctrl+U (Hue/Sat) and drag the saturation slider all the way to the right.
In a normal freq separation, the HF layer contains all the edges which are equal to and less than the chosen Radius and the colors that resided on both sides of those edges are intensified (contrast has increased along those edges). Conversely on the Low Frequency layer below, the same edges at the same radius have had contrast reduced (edges blurred). When the two layers are added together (function of the Linear Light Blend mode) the result is the original image. GB + HP = Original.
You are pretty safe using the healing brush on the HF layer because it tend to copy and blend in texture without taking all of the color. However you need to be careful using the Clone Stamp tool when trying to clone texture from an area of very dissimilar color because the clone tool will take all of the color as well and you can end up with a mess.
Regards, Murray

[BagLady]

@ Mistermonday

OOPS! Sorry for my mistake... I never noticed a significant amount of color in the high frequency layer. I was trying to explain things in a simple manor... In any case, thanks for correcting me Murray!

Regards, Baglady

[mistermonday]

Hi Baglady, no need to apologize. To almost everyone an HP layer looks like a colorless grayscale layer. It's not really obvious unless you know the math and mechanics going on inside PS.
Regards, Murray

[chillin]

Are there any disadvantages or cons, to placing a blank layer above the high frequency one and do the healing using a "current & below" sampler?

[mistermonday]

Hi Chillin, unfortunately you can't do that because the areas on the blank layer will represent the HF layer only and they can not blend in ADD (lineal light) mode with the GB layer.
The first concern I ever had with Frequency Separation was not having a fully re-editable layer in the HF layer. However in practice it has never been a problem and the speed and efficiency of that layer in the workflow far outweighs the risk.
Regards, Murray

[Flashtones]

Originally Posted by mistermonday

Hi Chillin, unfortunately you can't do that because the areas on the blank layer will represent the HF layer only and they can not blend in ADD (lineal light) mode with the GB layer.
The first concern I ever had with Frequency Separation was not having a fully re-editable layer in the HF layer. However in practice it has never been a problem and the speed and efficiency of that layer in the workflow far outweighs the risk.
Regards, Murray

Just to further MM's point:

I tend to amass far too many layers too quickly, but in doing so there's really no risk. Either, A) make the split from stamp visible layers, such that one could always regenerate the split again from the image below, or, B) dupe the HF layer and clone on the dupe, keeping a virgin copy turned off until one is satisfied the cloning is good, then trash the virgin, or, if need be blend the two with masks then merge them.

There's always an exit strategy if you use enough layers. ;-)

Also, this is from the original ModelMayhem thread, though I haven't tried it:

Just a quick tip to share.

If you are like me and don't like to clone/heal onto the original layers but instead onto a blank layer, this could present a problem when cloning/healing with the High Frequency layer. The only way to do this is to select "Current layer" for your sample. This would cause you to have to go back and forth from the HF layer to the blank layer each time you want to resample. Or, if you want to use "Current & below" for your sample, you can hide all layers below the HF layer but then you will only see the grey-ish HF layer which is not always the easiest to judge how the changes you are making will look in the final.

Cheer up, there's a way to change this to use "Current & below" for your sample and still see what the overall will look like so you can see what your changes are doing to the final. And it's very simple.

1. Hide all layers below the HF layer.
2. Make a copy of your LF layer and place it above the HF and HF heal/clone layers.
3. Change this copy to blend mode "Overlay".
4. Unhide this copy.

You should now see something that looks very close to the original. Close enough that you can determine what your cloning/healing on the HF is doing to the overall image. You can now also keep the HF heal/clone layer active and for sample use "Current & below" since the only layer visible below that is the HF layer.

You can also create an action that will hide all the layers below HF and unhide the LF copy and another action to do the opposite. Or, if you're like me, use a script that will do this for you as soon as you select the HF heal/clone layer and change it back when you select another layer.

I hope this is useful for some of you.

There may be other uses for this, also. I haven't explored the possibilities.

[chillin]

Thank you Murray, you got me thinking...again

Originally Posted by Flashtones

Just to further MM's point:

....or, if need be blend the two with masks then merge them.

There's always an exit strategy if you use enough layers. ;-)

:

Thank you Flashtones, this will work for me.

[Flashtones]

That's very clever, chillin, well done.

[chillin]

Originally Posted by Flashtones

That's very clever, chillin, well done.

Thanks Flashtones

[mistermonday]

Chiilin, you might be better off clipping a Blank layer to the HF linear light layer and use the eraser if it is needed. Another good option is to forget about a 3rd layer but instead create a layer group of the 2 split layers and just add a layer mask to the folder. Assuming you still have your underlying layers or BG, they will pass thru everywhere the mask has not been painted black. There are probably a couple of other ways. All that being said, I find in that for the type of healing done on the HF layer that you rarely make mistakes and when you do they are easy to fix. Its nice having a safety net for everything you do in a retouch but if you work on large 16 bit files, every additional full pixel layer bloats the file and slows down the processing considerably.
Regards, Murray

[chillin]

Yes you're right, the file gets considerably larger, but I don't like the thought of not having easy way out

Thanks Murray