Smudging when using Frequency Separation

[CKAJCA]

I am still very new to retouching but have found Frequency Separation for use when retouching to be extremely useful. I still have a problem that I do not know how to overcome. I am still getting smudging when cloning with a hard edge brush on the High layer. An example would be when trying to remove black flyaway hairs against skin at the point where the flyaway hair meets the main body of the hair. This is obviously an area of high contrast and even with the clone tool set to hard it is still smudging a bit. I suspect this is something to do with the fact that there is still a small amount of colour on the High layer.
Can anyone tell me if there is a way around this or how best to handle it.

Colin

[k4stn]

Hi Colin

To remove color from high pass. Image, ajustmeant, hue/saturation and put saturation all the way down to -100

Shane

[Der_W]

Please don't desaturate your image after the split! You might want to keep all the color information on the lowpass (search for asymmetric frequency separation in that case), but I rather suspect that the smudging comes from you working near hard edges. Try using a different filter that largely respects edges to create your lowpass and see how that works. Median or Surface Blur might work for this.
Also when healing near edges, always sample the edge itself and work along it, that way it won't blur that much.

[mistermonday]

Originally Posted by CKAJCA

I am still very new to retouching but have found Frequency Separation for use when retouching to be extremely useful. I still have a problem that I do not know how to overcome. I am still getting smudging when cloning with a hard edge brush on the High layer. An example would be when trying to remove black flyaway hairs against skin at the point where the flyaway hair meets the main body of the hair. This is obviously an area of high contrast and even with the clone tool set to hard it is still smudging a bit. I suspect this is something to do with the fact that there is still a small amount of colour on the High layer.
Can anyone tell me if there is a way around this or how best to handle it.

Colin

Colin, while it does not look like it, there is a significant amount of color on the HF layer. Everything that is not gray is actually an edge of your chosen radius where the color contrast has been greatly intensified. By performing an asymetric frequency separation you will remove all of the color from the HF layer leaving all of the color on the LF layer but still retaining the integrity of the split layer. Here is a link describing how to make one:

http://www.retouchpro.com/forums/pho...on-method.html

Regards, Murray

[CKAJCA]

Originally Posted by mistermonday

Colin, while it does not look like it, there is a significant amount of color on the HF layer. Everything that is not gray is actually an edge of your chosen radius where the color contrast has been greatly intensified. By performing an asymetric frequency separation you will remove all of the color from the HF layer leaving all of the color on the LF layer but still retaining the integrity of the split layer. Here is a link describing how to make one:

http://www.retouchpro.com/forums/pho...on-method.html

Regards, Murray

Hi Murray, that is amazing and works fantastically well, very much appreciated for sharing that technique. You are a star!

[CKAJCA]

I seem to have a problem when using the clone stamp tool on the HF layer. When I clone I have noticed that although the item I am cloning onto goes (for example a flyaway hair on the skin) I am left with an orange/saturated skin looking colour. Is there a way to overcome this?

Colin

[mistermonday]

Originally Posted by CKAJCA

I seem to have a problem when using the clone stamp tool on the HF layer. When I clone I have noticed that although the item I am cloning onto goes (for example a flyaway hair on the skin) I am left with an orange/saturated skin looking colour. Is there a way to overcome this?

Colin

Yes, make sure you have the tool set to sample "Current Layer" only otherwise you will get weird colors. All the texture is on the HF layer and you do not want to sample color from the layers below.
Regards, Murray

[CKAJCA]

Originally Posted by mistermonday

Yes, make sure you have the tool set to sample "Current Layer" only otherwise you will get weird colors. All the texture is on the HF layer and you do not want to sample color from the layers below.
Regards, Murray

Hi Murray
I am aware of that problem and solution because I made that mistake shortly after I started to use Frequency Separation ;-) This is a much more subtle orange/ saturated skin colour and appears everywhere I clone onto. I thought I could fix it by working on the Low layer but often I cannot get rid of it. Any ideas?

Also as I am still new to this Frequency Separation technique, can you advise me how best to set the level of Gaussian Blur on the Low layer. Any guidance would be appreciated.

Colin

[mistermonday]

Originally Posted by CKAJCA

Hi Murray
I am aware of that problem and solution because I made that mistake shortly after I started to use Frequency Separation ;-) This is a much more subtle orange/ saturated skin colour and appears everywhere I clone onto. I thought I could fix it by working on the Low layer but often I cannot get rid of it. Any ideas?

Also as I am still new to this Frequency Separation technique, can you advise me how best to set the level of Gaussian Blur on the Low layer. Any guidance would be appreciated.

Colin

It would be best if you could post a sample image and/or area affected (before and after).
Choosing the correct radius is important. Further more a single radius is not always appropriate to treat an entire image. Often you need to make 2 or 3 separations, each at a different radius.
Regards, Murray

[CKAJCA]

Originally Posted by mistermonday

It would be best if you could post a sample image and/or area affected (before and after).
Choosing the correct radius is important. Further more a single radius is not always appropriate to treat an entire image. Often you need to make 2 or 3 separations, each at a different radius.
Regards, Murray

Hi Murray
Here are before and after shots of approx the same area. I hope you can see the orange/saturated skin spots where I have cloned over?

[CKAJCA]

Here is a larger version

[CKAJCA]

Apart from the orange/saturated skin colour I can also detect a slight grey colour as well!

[mistermonday]

Can you please also post a larger sample of the before. Your radius may not be optimized.
Regards, Murray

[CKAJCA]

Here is a larger version of the Before

[mistermonday]

You do not need to use the Asych Freq Sep on this image. The regular freq sep works fine. I tried it on a nuber of areas and do not get any adverse affects or intensified color. Moreover much of your Before image is a good candidate for conventional Dodge & Burn.
What Radius did you use for the separation. I need to go offline for a few hours. I will check back later this afternoon.
Regards, Murray

[CKAJCA]

Originally Posted by mistermonday

You do not need to use the Asych Freq Sep on this image. The regular freq sep works fine. I tried it on a nuber of areas and do not get any adverse affects or intensified color. Moreover much of your Before image is a good candidate for conventional Dodge & Burn.
What Radius did you use for the separation. I need to go offline for a few hours. I will check back later this afternoon.
Regards, Murray

I only used 3 on the Gaussian blur!

[KTG]

Murray,
I have found that trying to get the "right" frequency to be challenging sometimes. The rule of thumb, I believe, was to blur until the detail you want to save is blurred out. Is that how you determine the frequency choice? Also, in IHP, how do you make the choice for High Pass radius. I blur the area that I want to effect and use that radius for the High Pass radius. Of course, that will vary in the image. And often I will have two or three layers of IHP.

k

[mistermonday]

Originally Posted by CKAJCA

I only used 3 on the Gaussian blur!

I do not know the size / resolution of your original, but I would bet that 3 px radius is way to small. I would expect it to be 10 or higher.
Regards, Murray

[mistermonday]

Originally Posted by KTG

Murray,
I have found that trying to get the "right" frequency to be challenging sometimes. The rule of thumb, I believe, was to blur until the detail you want to save is blurred out. Is that how you determine the frequency choice? Also, in IHP, how do you make the choice for High Pass radius. I blur the area that I want to effect and use that radius for the High Pass radius. Of course, that will vary in the image. And often I will have two or three layers of IHP.

k

Yes, blur until the detail you wish to save is blurred. Actually it's the detail you wish to isolate because you may also want to repair it (with clone tool), or accentuate it or other.
For IHP, you are trying to create a 3rd band of edges whose radii cover the range between the top end of the fine detail edges and the lower end of the large radii edges. Because everything is inverted to obtain this band, the HF radius value is the one which would be what you would use for Gaussian Blur and the Gaussian Blur radius would be what you would normally use for High Pass. I have really two answers to your question regarding IHP because there are two primary uses of it.
1. If you are using IHP in a retouch to subtly soften harsh edge transitions of light to dark / dark to light, this is usually done at high radius. Your High Pass Radius would typically be 50 / 60/ 70 / 80 px. The GB radius would typically be 1/3 of the HP value. This is a reasonably good rule of thumb and pretty reliable.
2. If you are using IHP as a Degrunge technique to blur away small / medium textures, then the rule of 1/3 is wrong most of the time. Instead what you should do is do a Gaussian Blur backing down the radius until the detail you wish to keep almost reappears. You can see an example at an old tutorial at RP here: http://retouchpro.com/tutorials/?m=show&id=213
Today, the plain old High Pass filter has fallen out of favor because it is inaccurate and has been replaced with the Apply Image technique but the concept of IHP is still a variant of Degrunge. The principles and math are still the same.
Regards, Murray

[kav]

Originally Posted by mistermonday

Colin, while it does not look like it, there is a significant amount of color on the HF layer. Everything that is not gray is actually an edge of your chosen radius where the color contrast has been greatly intensified. By performing an asymetric frequency separation you will remove all of the color from the HF layer leaving all of the color on the LF layer but still retaining the integrity of the split layer. Here is a link describing how to make one:

http://www.retouchpro.com/forums/pho...on-method.html

Regards, Murray

Murray please do not take this personally, but you're making that needlessly complex. The technique being talked about here is no less inherently destructive than any other method of retouching. By duplicating so much stuff you're inflating your file size without getting much for it. You really do not need to split your image data into parts to achieve good results.

Originally Posted by CKAJCA

I am still very new to retouching but have found Frequency Separation for use when retouching to be extremely useful. I still have a problem that I do not know how to overcome. I am still getting smudging when cloning with a hard edge brush on the High layer. An example would be when trying to remove black flyaway hairs against skin at the point where the flyaway hair meets the main body of the hair. This is obviously an area of high contrast and even with the clone tool set to hard it is still smudging a bit. I suspect this is something to do with the fact that there is still a small amount of colour on the High layer.
Can anyone tell me if there is a way around this or how best to handle it.

Colin

Colin you're best off forgetting you ever read such a technique. It's quite easy to do this without the "split" technique. You can use clone lighten mode or a blank lighten layer to clone these out. There are also a couple ways to dodge them out for the really annoying ones. You just have to do it carefully, but anything that's giving you smudgy results is a waste of time. I went through a lot of these internet distributed techniques a decade or so ago. Trying to master cheap techniques is basically doing yourself a disservice. That one has one redeeming quality. You can extract relatively clean texture if you have to rebuild an area with really flat texture. For that it might work better than extracting via high pass.

[mistermonday]

Originally Posted by kav

Murray please do not take this personally, but you're making that needlessly complex. The technique being talked about here is no less inherently destructive than any other method of retouching. By duplicating so much stuff you're inflating your file size without getting much for it. You really do not need to split your image data into parts to achieve good results.

Kav, I was just responding to the questions, not commenting on usage. Like any other tool or technique in retouching, this one has it's advantages and disadvantages. It will add file size but so does every additional layer. It can add complexity but so do other methods. It can save you a large amount of time on some images but not on others. It's a trade off and you need to make an assessment.

Regards, Murray

[kav]

Originally Posted by mistermonday

Kav, I was just responding to the questions, not commenting on usage. Like any other tool or technique in retouching, this one has it's advantages and disadvantages. It will add file size but so does every additional layer. It can add complexity but so do other methods. It can save you a large amount of time on some images but not on others. It's a trade off and you need to make an assessment.

Regards, Murray

I think I was just grumpy yesterday. I really need to learn not to post when I'm in a bad mood. It could be fun to compare. If anyone has one where that technique made it considerably easier, show me and I'll see how I'd replicate it a different way to compare. I've tried out dozens of things in attempts to increase efficiency or prevent things like flat spots. In the end keeping it to relatively few layers for the cleanup work helps immensely. I don't do much directly on image data. If it's something really tiny that can't be dodged out I'll clone it. If it's more than a couple pixels wide I'll rebuild it instead. Then it's just a few lighten/darken layers. I can't stand spreading it among too many layers, as it can have a tendency to mess with the color/local contrast.

Splitting things up several times really does look tedious to me, even if I don't like the way I worded that previous post.