Lady in red

[William W]

. . .there are many paths open to all of us. . . . I think it is often very useful to explain to people why an image violates one or another guideline, particularly if one can explain why that guideline often makes sense.

You get it. Thank you. Whether it was Javier's intent or not doesn't matter - it is in the image and it was noticed by quite a few, and hopefully, explained the 'why' with some degree of understanding.

WW

[xpatUSA]

Ted - Thank your sharing your opinion.

1. Using a good MUA negates the need for some advanced retouching and is often the best way to photograph a model. So it is indeed relevant when discussing how to retouch. Less is more.

OK.

...

2. I've never suggested that there is no place for frequency separation as a technique, including retouching portraits. There are times where is is probably the best approach, but those are generally the exception, not the rule in high end work. With frequency separation we destroy the natural skin textures and replace them with simulated skin textures. We are effectively creating "fake" skin ...

Sorry but I completely disagree with that all-encompassing statement. No destruction or simulation is involved beyond that found in "high-end" editing, IMO.

[DanK]

I've only dabbled in frequency separation, but I'm puzzled. My understanding is that the point of frequency separation is that one can make edits to either low-frequency or high-frequency information without affecting the other. For example, one can remove a blemish without accidentally changing color. Wouldn't that be more natural rather than less? It seems to me that the issue with frequency separation, as with any other editing, is to avoid being heavy-handed and to try to avoid inadvertent collateral damage.

[Manfred M]

Dan - you are completely right Dan. Most of the problems we see with frequency separation are application issues. Low frequency are used for large areas of skin colour and are handy for repairing those areas. High frequency is used to take care of fine details like pores. Getting the balance right between the two is what makes the skin look right or not.

What we often see is that the low frequency layer is given too much precedence and we get plastic skin. When the high frequency layer is given too much precedence the skin looks like sandpaper. Getting the mix just right is important.

Let's also separate what high end retouching is from "normal" retouching. High end is the stuff you will see in high end publications, high end custom portraits (generally large prints) and especially in advertising material from "A" list companies; the national and international brands most of us would recognize.

The "normal" stuff, especially out of photographers that deal with retail (members of the public) clients, like wedding and portrait photographers would either use actions in Photoshop (based on Frequency Separation) or use a plugin. The clients would likely not see the difference. I use Imagnomic Portraiture 3 when trying to save time in "mass production". The only time I use anything more extreme in my work is either when I am doing a portraiture course or preparing large prints for competitions.

[William W]

. . . Let's also separate what high end retouching is from "normal" retouching.

> High end is the stuff you will see in high end publications, high end custom portraits etc
> The "normal" stuff, especially out of photographers that deal with retail (members of the public) clients, like wedding and portrait photographers would either use actions in Photoshop (based on Frequency Separation) or use a plugin.

Agree. Good point.

Plugins used often in the second category.

I think that plugins are used by many "enthusiastic" Photographers, moreover the marketing and advertizing for plugins is directed to that group.

WW

[xpatUSA]

Dan - you are completely right Dan. Most of the problems we see with frequency separation are application issues. Low frequency are used for large areas of skin colour and are handy for repairing those areas. High frequency is used to take care of fine details like pores. Getting the balance right between the two is what makes the skin look right or not.

Good to see that 'frequency separation' can actually make skin look right ...

What we often see is that the low frequency layer is given too much precedence and we get plastic skin. When the high frequency layer is given too much precedence the skin looks like sandpaper. Getting the mix just right is important.

... in spite of these oft-repeated caveats.

[Manfred M]

Agree. Good point.

Plugins used often in the second category.

I think that plugins are used by many "enthusiastic" Photographers, moreover the marketing and advertizing for plugins is directed to that group.

WW

One of my teachers / mentors is an "A" list photographers with clients like Chanel, ;L'Oreal, Toyota. Chrysler, etc, etc. He has just completed a 4 - 5 month assignment photographing four coins for the Royal Canadian Mint. He had set up a studio at their location (security reasons), so the shooting conditions were close to ideal. He uses a high end retoucher and just received the invoice; there were more than 140 hours of retouching that went into the final delivered images (I don't know the exact number, but suspect we are looking at around a dozen or so shots). That suggests in the order of 10 hours per image of retouching, using someone who does this full-time for a living, with a polished workflow and a lot more speed than I have.

As for plugins are used by every wedding and studio portrait photographer I have spoken with uses them as a way of speeding up workflow. When time is money and you are delivering 1000+ images at a wedding and reception, speed is the driving force in workflow.

[Manfred M]

Good to see that 'frequency separation' can actually make skin look right ...
.

Depends on who is using the tool and their competence level. For run of the mill images being delivered to members of the public, that is all that is needed. They wouldn't notice the difference as long as the retouch has been well done. If it hasn't....

As I said before, the high end stuff is often just dodging and burning.

[xpatUSA]

Originally Posted by xpatUSA Good to see that 'frequency separation' can actually make skin look right ...

Depends on who is using the tool and their competence level. For run of the mill images being delivered to members of the public, that is all that is needed. They wouldn't notice the difference as long as the retouch has been well done. If it hasn't....

Manfred, you keep applying adverse "spin" to the frequency separation method: e.g. "run-of-the-mill", "wouldn't notice the difference", "if it hasn't", et prev. Are we to understand from your comments that it would be impossible for a high-end re-toucher using that method to deliver a satisfactory image to a commercial client?

As I said before, the high end stuff is often just dodging and burning.

As before, this method mentioned with no caveats as to competence ...

I keep coming back because of the way your responses are worded - always favoring the one less than other.

May I suggest:

Frequency separation can render skin correctly.

High-end re-touching can render skin correctly.

Fair?

I guess I'll get out of your hair now ...

[Manfred M]

Let me demonstrate three different techniques. Open the images in Lightbox mode by clicking on one of the images and step through the various approaches.


1. This is a crop at 100% of the original image. Model has makeup that has been reasonably well appiied.






2. This is a retouch with Frequency Separation






3. This is a retouch with a plug-in. I've used the relatively high end (from a cost standpoint) Imagenomic Protraiture 3






4. Dodging & Burning

[Manfred M]

Originally Posted by xpatUSA Good to see that 'frequency separation' can actually make skin look right ...

Manfred, you keep applying adverse "spin" to the frequency separation method: e.g. "run-of-the-mill", "wouldn't notice the difference", "if it hasn't", et prev. Are we to understand from your comments that it would be impossible for a high-end re-toucher using that method to deliver a satisfactory image to a commercial client?



As before, this method mentioned with no caveats as to competence ...

I keep coming back because of the way your responses are worded - always favoring the one less than other.

May I suggest:

Frequency separation can render skin correctly.

High-end re-touching can render skin correctly.

Fair?

I guess I'll get out of your hair now ...

Ted - I have both a philosophical and practical issue with the frequency separation method, because in my experience it results in the least natural look of the three methods.

In the low frequency layer, the various skin tones merge and the skin tends to look unrealistic. With the high frequency layer, it is all too easy to introduce small directional changes in the size / direction of the high frequency features so once again, the results can look a little "off". There is a role for the tool, but I only use it in places where the skin is very patchy. I prefer to use other tools; dodging, burning and changing the saturation of individual colour channels. Sometimes the skin (or poor makeup application) can be so bad, that it is the best option in limited areas.

[xpatUSA]

Ted - I have both a philosophical and practical issue with the frequency separation method, because in my experience it results in the least natural look of the three methods.

In the low frequency layer, the various skin tones merge and the skin tends to look unrealistic. With the high frequency layer, it is all too easy to introduce small directional changes in the size / direction of the high frequency features so once again, the results can look a little "off".

Thank you Manfred for the further discussion and the demo images in the previous post. I have never used the PS frequency separation method, so perhaps I should not be supporting it quite so vehemently. I myself use wavelet processing with 5 or more detail levels (frequencies) with of course a residual image to play with, so perhaps more sophisticated than Adobe's offering.

I also admit to never having shot models of any kind and am therefore unable to quote any experience, unlike your good self.

There is a role for the tool, but I only use it in places where the skin is very patchy. I prefer to use other tools; dodging, burning and changing the saturation of individual colour channels. Sometimes the skin (or poor makeup application) can be so bad, that it is the best option in limited areas.

[Manfred M]

I also admit to never having shot models of any kind and am therefore unable to quote any experience.

Ted - that is part of the problem in trying to explain. Unless you work in the genre, you will never see the nuances as these take some time to learn and understand. Most people will never see the difference between the various tools. People that do a lot of portraiture do.

Unfortunately, "experts" do and can spot the differences between a strong and a weak skin retouch.

[Round Tuit]

Let me demonstrate three different techniques. Open the images in Lightbox mode by clicking on one of the images and step through the various approaches.


1. This is a crop at 100% of the original image. Model has makeup that has been reasonably well appiied.






2. This is a retouch with Frequency Separation






3. This is a retouch with a plug-in. I've used the relatively high end (from a cost standpoint) Imagenomic Protraiture 3






4. Dodging & Burning

Manfred - Thank you for taking the time to create this demonstration of three skin retouching methods. I found it very to be a very good learning tool. I loaded the four pictures on 4 layers in Photoshop so that I could do an A B comparison between each pair. The areas that you retouched really stood out in the comparison.

I am one of the "people" who would not be able to tell the differences between them taken in isolation. Looking at them side by side though I find the "dodge & burn" has the most realistic skin texture. To my untrained eye, the "plug-in" one has the most artificial looking skin with the "frequency separation" being somewhere in between.

Given my limited understanding of the frequency separation technique as described in this post, I find it puzzling that the retouching changed the texture of the skin. So I set up the method in Photoshop to try to replicate your result. I used a 5 px Gaussian blur on the low frequency layer and used the apply image to create the high frequency layer. I used two curve adjustment layers between the two; one for burning and one for dodging.

I retouched the following areas:
1. Healed most of the white specs on her right cheek and left eye,
2. Dodged the dark skin patch at the end of her right eyebrow,
3. Partially dodged the dark areas around her eyes, and finally
4. Burned the ridge of her nose.

This is the result that I got:



The extent of the retouching that I did may be different that what you did but to my untrained eye, the skin texture looks very similar to your "dodge and burn" version and definitely not as plastic as your "frequency separation" version. I will be the first one to admit that I am a neophyte when it comes to portraiture and even more so for "high end retouching" but it seems to me that the technique has potential to be very useful in retouching textured areas.

[xpatUSA]

Manfred - Thank you for taking the time to create this demonstration of three skin retouching methods. I found it very to be a very good learning tool. I loaded the four pictures on 4 layers in Photoshop so that I could do an A B comparison between each pair. The areas that you retouched really stood out in the comparison.

I am one of the "people" who would not be able to tell the differences between them taken in isolation. Looking at them side by side though I find the "dodge & burn" has the most realistic skin texture. To my untrained eye, the "plug-in" one has the most artificial looking skin with the "frequency separation" being somewhere in between.

Thanks for the idea of comparing in layers, André. As Manfred pointed out, I don't have "the eye" either.

So instead, I compared the three methods to the original by setting their layer modes to "difference". The plugin method showed a really dark homogenous screen, indicating very little difference, pixel-wise.

However, frequency separation versus dodging and burning was quite telling:



Both images' contrast increased to "64" (FastStone metric) so as to emphasize the differences from the original.

The frequency separation method looks harsh in places (yellow outline) and Manfred's work on texture came out really well (red outline).

I am left wondering whether the simple two-frequency Photoshop method could be bettered by, say, 5-level wavelets+residual ...

[Manfred M]

Manfred - Thank you for taking the time to create this demonstration of three skin retouching methods. I found it very to be a very good learning tool. I loaded the four pictures on 4 layers in Photoshop so that I could do an A B comparison between each pair. The areas that you retouched really stood out in the comparison.

I am one of the "people" who would not be able to tell the differences between them taken in isolation. Looking at them side by side though I find the "dodge & burn" has the most realistic skin texture. To my untrained eye, the "plug-in" one has the most artificial looking skin with the "frequency separation" being somewhere in between.

Given my limited understanding of the frequency separation technique as described in this post, I find it puzzling that the retouching changed the texture of the skin. So I set up the method in Photoshop to try to replicate your result. I used a 5 px Gaussian blur on the low frequency layer and used the apply image to create the high frequency layer. I used two curve adjustment layers between the two; one for burning and one for dodging.

I retouched the following areas:
1. Healed most of the white specs on her right cheek and left eye,
2. Dodged the dark skin patch at the end of her right eyebrow,
3. Partially dodged the dark areas around her eyes, and finally
4. Burned the ridge of her nose.

This is the result that I got:



The extent of the retouching that I did may be different that what you did but to my untrained eye, the skin texture looks very similar to your "dodge and burn" version and definitely not as plastic as your "frequency separation" version. I will be the first one to admit that I am a neophyte when it comes to portraiture and even more so for "high end retouching" but it seems to me that the technique has potential to be very useful in retouching textured areas.

Andre - When I look at your edit, it looks slightly different than mine, which is okay, as we likely used slightly different settings and work flows in our editing approach. I notice the same issue in both; the areas under the eyes don't look like that in real life, i.e. the post processing work is not 100% believable.

I selected the closeup of the model's eyes, because that is one area that I can often pick up that frequency separation has been used. The fine wrinkle lines exist in everyone's eyes one they are no longer young children. In my view the key in retouching eyes is to make them look both realistic and somewhat more youthful. The difference, in adult eyes, are the quantity and depth of the wrinkles. These result in highlights and shadows of those facial features and by reducing their intensity, rather than in eliminating them we get a more natural and more "believable" look. As we are dealing with highlights and shadows, that's why dodging and burning are so effective.

That being said, most people are not going to notice the difference, which is why frequency separation is so popular. They are going to notice the lack of wrinkles and If you get someone who is, they will pick up on it, depending on the image itself.

[xpatUSA]

I selected the closeup of the model's eyes, because that is one area that I can often pick up that frequency separation has been used. The fine wrinkle lines exist in everyone's eyes one they are no longer young children. In my view the key in retouching eyes is to make them look both realistic and somewhat more youthful. The difference, in adult eyes, are the quantity and depth of the wrinkles. These result in highlights and shadows of those facial features and by reducing their intensity, rather than in eliminating them we get a more natural and more "believable" look.

Thanks for the tip, Manfred. Having fun in wavelets as we speak ...

To close that one out, I never achieved perfection with global wavelet processing.

[Round Tuit]

Andre - When I look at your edit, it looks slightly different than mine, which is okay, as we likely used slightly different settings and work flows in our editing approach...

Manfred - You are very generous to say that. The difference is that you know what you are doing and I don't. This is the first decent portrait that I have ever retouched so I tried to approximate the look of your edits. The results were close enough to convince me to pursue the frequency separation approach but not with portraits. I will stick with flowers and landscapes for a while.