Guide to Frequency Separation
Frequency separation is a very effective post-production technique that allows us to work selectively on our images.
It’s commonly used for portrait retouching as it allows us to remove skin imperfections and to obtain a final effect with soft skin without removing the skin texture.
In this article, we’ll see what frequency separation is all about and how to use it in a simple way.
What is Frequency Separation?
Each of our digital images consists of its most fundamental elements: pixels. And each pixel is associated with fundamental information to describe our image.
If taken individually the pixels can’t tell us much… in fact, they’re only colored squares. But when they’re put together and related to each other, not only do they allow us to see our final image, but they allow us to make considerations about how they relate to each other.
There are many ways to analyze these mutual relationships, and one of the most used is frequency separation.
In essence, it’s possible to break down our image into spatial frequencies: a technical and complicated word to say that we can make divisions according to the level of detail contained within different areas, and then act selectively on these zones.
Although more formally we should consider three levels of frequency, we can easily think of dividing our image into high frequencies and low frequencies.
The high frequencies are those with very fine details – in portraits, for example, the hair, skin pores, small imperfections… everything you see when you look at the image very close up.
The low frequencies are those that instead enclose volume information, such as tonal transitions, lights, and shadows... everything you see in the image when you look at it from further away.
An example that’s often done, and that I also love to do to explain this concept, is the famous optical illusion “Marilyn Einstein” created by MIT.
If you look at this image from a distance (try to move 5 meters away from the screen), you’ll see without too much difficulty the image of Marilyn Monroe.
The closer you get to the screen, the more the image of Albert Einstein replaces that of Marilyn.
How is this possible? Through frequency separation!
This image was created by superimposing the High Frequencies (fine detail) of a picture of Albert Einstein onto the Low Frequencies (volume information) of a picture of Marilyn Monroe.
How to Use Frequency Separation in Photoshop
Credit: Vojta Dzubák
There are many software applications that allow you to apply the frequency separation technique, but the most popular is undoubtedly Adobe Photoshop. Therefore, we’ll cover the steps for retouching in Photoshop using frequency separation.
Frequency separation is basically done in three steps:
- Creation of frequency layers
- Setting frequency layers
- Application of corrections
Step 1: Creation of Frequency Layers
First, let’s duplicate our background layer (which is the image we want to modify) and create two independent layers.
To do this we just use Ctrl + J or ⌘ + J on the background layer. Alternatively, right-click on the background layer and press “Duplicate Layer”.
This operation must be done twice as we have two frequency layers to handle.
At this point, we rename the two layers as needed. You can rename them according to the corrections you want to apply, or more simply, as I normally do, as “High frequencies” and “Low frequencies”.
If the frequency separation adjustments are part of a more complex workflow with many other layers of adjustment, I suggest you use the “Group from layers” function to group them into a folder. To do this, simply select the two layers you just created and press Ctrl + G or ⌘ + G.
As you can see, in my example I’ve named the group of layers “Frequency Separation” to keep things clear and organized.
Step 2: Setting Frequency Layers
Once the two layers are created, we have to set them properly.
Let’s start with the low frequency layer. Make sure you have it selected, taking care to remove the visibility of the top layer (the high frequency layer – you can press on the eye icon on the left of the layer name).
Now let’s apply a Gaussian Blur filter to our low frequency layer.
Go to Filters -> Blur -> Gaussian Blur.
This will bring up a dialog box where you can set the intensity of the filter. I suggest you stay in the range between 6 and 10.
The preview will show your image completely blurred: don’t worry, this is normal because we’ll act selectively only afterwards.
At this point, we need to prepare the high frequency layer. To do so, first we reactivate the visibility of the top layer we had hidden in the previous step.
Now select Image -> Apply Image and a new window will open.
Here, we’ll be able to combine the information of the low frequency layer with the high frequency layer we’re working on.
In the dialog box, enter the settings you see in the image below:
Here’s a summary of the settings:
- Layer: Low frequencies
- Channel: RGB
- Blending: Subtract
- Opacity 100%
- Scale: 2
- Offset: 128
The preview of your photo will result in something very strange: a gray image with some details in evidence. Don’t worry, it’s perfectly normal!
Basically, you’re asking Photoshop to merge the two layers of frequencies you created, subtracting one from the other.
Scale and Offset are two basic parameters as they tell Photoshop how to work. A higher value of Scale would make the effect too pronounced, while a different value of Offset would compromise the brightness of the image.
These settings are optimal if you’re working with an 8-bit image. If you’re working with 16 bits, to get a correct result you should set the values as in this image:
As you can see, in this case we’ve actually reversed the selection and applied an “Add” blending method, setting an Offset equal to 0 with the same Scale.
Again, the preview will look strange (specifically white with some visible detail), but we don’t have to worry.
Once you’ve entered the above parameters, press OK.
At this point, your image is still unrecognizable because we still have to do one last step.
Taking care to keep the high frequency layer selected, switch the blend method of the layer to Linear Light (instead of Normal) through the drop-down menu.
Now you’ve correctly separated your image by frequency!
As you’ll notice, your image will be identical to the starting point. This is absolutely correct because the sum of the parts of our decomposition must add up to the original image.
This is the basis for any skin retouching action, as you can now really act selectively on your image.
Step 3: Application of Corrections
The third and final step is obviously to apply corrective actions depending on the result we want to achieve.
A first adjustment can be simply varying the opacity values of our two frequency layers: by modulating these values we can give more weight to the low or high frequencies, softening the image or making it harder.
This is not an optimal adjustment because we’re going to act globally on our image, losing all the potential of the frequency separation made so far.
The ideal is therefore to work with special tools depending on what we want to achieve on the frequency layer that best lends itself to the correction… and it’s exactly for this reason that we made a frequency separation!
In the next section, we’ll look at some ways to make the most of our high and low frequency layers.
Tips for Perfecting Frequency Separation in Photoshop
As mentioned earlier, creating the two frequency separation layers is the basis for accurate selective corrections.
There are no limits to the tools you can use and the various techniques you can apply: the only thing to do is to ask yourself beforehand which level your specific correction is best applied to.
I’ll now give you some practical examples of how to improve your images using Photoshop frequency separation.
1. Skin Softening
Undoubtedly one of the most pleasant and sought-after effects in portrait post-production is the softening of the skin.
When you try to work on an image where no frequency sepration has been done, the result is often disappointing and unrealistic. The skin texture and other fine details are lost, and you end up with the “porcelain doll” effect.
Our eye likes details, and this is where the frequency separation comes in handy.
A trick to soften the skin is to act on the low frequency layer only by applying a Gaussian Blur filter where we really need it.
To do this, first select an area of the face with the Lasso tool and set a Feather between 10px and 20px. Selecting the Feather allows us to soften the transition between the area we’re going to soften and the area we’re not going to touch up.
After that, let’s select Filter -> Blur -> Gaussian Blur and set a value between 10px and 25px.
As you can already see from the preview, the skin smoothes and softens, but without losing detail: this is due to the fact that we’re softening the low frequency layer without contaminating the detail contained in the high frequency layer.
2. Skin Imperfection Removal
The technique of frequency separation is also particularly useful for the removal of imperfections such as blemishes from the skin.
There are many different ways to do this type of skin retouching in Photoshop, but once again working on the correct frequency layer helps us to have a more natural result.
As with skin softening, I suggest you remove any blemishes on the low frequency layer.
To do this simply, I use the Patch tool. This can be activated either through the Tools menu or by simply pressing the J key.
All you need to do is go and select the impurity and drag it to an adjacent area.
For a more natural result, I suggest you make your selection with a little margin around the blemishes.
Also, I recommend dragging the selected impurity to a zone with about the same brightness: this allows Photoshop to better calculate the correction and to have a more homogeneous feeling.
You can continue to repeat the operation on all skin imperfections and blemishes until you get the skin as clear as you want.
As you’ll see, the blemishes are removed but the detail in the skin texture is not altered because we’re working again on the low frequency layer and not the high frequency one.
Note that while I’ve used the Patch tool in this example, you can use whatever blemish-removal process works best for you. For example, the clone stamp tool, spot healing or healing brush tool can all work.
The result of keeping the skin texture while smoothing out the blemishes will be the same.
3. Highlighting Details
Frequency separation is not only done to take advantage of low frequencies. You can also use it to bring out some details that are not necessarily related to your skin.
One example relates to the eyes.
When we look at a face, we’re normally led to look at it in one of the eyes – and the same thing happens with a portrait image. Since we normally focus on the eye, it’s a detail that we expect to see perfectly sharp.
There are several more or less advanced techniques to emphasize this detail of sharpness, but now I will show you one that’s very simple and works well in most cases.
Start by creating a white mask in the low frequency layer. To do this, hold down the Alt (Windows) or Option (Mac) key and press “Add Layer Mask” on the small bar at the bottom of the layers panel.
In this way, we’ll create a white mask without any adjustments applied.
At this point, leaving this white mask selected, we select a black brush set with very low hardness and with a size appropriate to the area we’re going to retouch.
Let’s now apply the brush to the area of interest, so in our case the eye.
As you will see, the brush strokes remove a kind of opacity from the eye, making it brighter and more defined.
Basically, we masked the low frequencies on the eye, letting the high frequencies stand out.
This trick also works particularly well with hairs (when we have them in focus) and with eyebrows, for example – ideal for glamour photography and anything where the model’s features need to be delicately highlighted.
4. Shadow Management
Another interesting application of the frequency separation technique is related to localized shadow management.
In a portrait, depending on the available lighting we do not always have the possibility to manage light and shadow as we’d like.
If we simply opened up the shadows in our base image, we would once again compromise both detail and color. By doing so on the divided frequency layers, on the other hand, we can maintain maximum control.
To do so, let’s first select the low frequency layer.
Now we select the Dodge tool in the tools bar.
To get the best result, first select a soft brush. In the Range drop-down menu, select Shadows and set the Exposure to around 15%.
Setting a brush of an optimal size for the retouch we want to do, let’s brush on the shadow areas we want to lighten.
As you can see, shadows are recovered but skin texture is not compromised once again.
Be careful not to exaggerate as we always want to maintain a clean and realistic result.
As you can see, the technique of frequency separation is actually very easy to apply. The key benefit is that it allows us to obtain, in a simple way, really selective and professional corrections to our images.
In portrait photography, it’s almost indispensable to make corrections on the skin that don’t alter reality and that act locally and not globally.
Once we separate the high and low frequencies in our image, we can act only on the layer that contains the detail that interests us.
If you’ve already mastered retouching using the spot healing brush or clone stamp tool, separating out the frequencies first will take your edits to the next level and allow you to retain far more texture and detail.
In this article, we’ve seen some simple examples using the basic tools that Photoshop provides us. But remember that once the frequencies are separated, we can apply any type of correction, even by inserting complex masks managed with additional layers.
In short – now there’s nothing left to do but experiment!