Frequency separation

In portraiture photography, probably one of the most widespread and important processing techniques is frequency separation. This technique allows us to separate the picture information into two layers, one that contains the color information and other that contains the texture information. Once the separation is achieved, you can modify, for example, the color of the skin without altering the texture it has. And on the contrary, you can modify little imperfections as blemishes, scars or pimples without having to take care of the color of the skin. It is an advanced technique that requires some practice to exploit everything it offers, but the truth is that, once dominated, allows for a great flexibility on processing and spectacular effects correcting the face of the model.


Why two frequencies?

When we analyze a picture, there are two kinds of information you can focus on. On one side you have colors. Colors usually are uniform areas that remain the same or contain very little variation along the image. If you imagine the picture of a leaf flying against the sky, you can coarsely describe it as a huge blue background with a green spot in the middle (you can add a second white irregular spot if you want to consider a cloud also in the picture). Perhaps the blue background is not uniform, but it changes softly along its area. The same happens with the leaf and the cloud, might be irregular, but are well confined and the hue and luminosity is usually similar along all the spot.

Of course, the former description is a bit too much coarse to describe a great picture, but it is obvious that contains a lot of information about the picture; vague and “soft” information, but useful to describe it anyways. That is what is called a low frequency. If you imagine music waves, they are composed of many waves of different frequencies. If you look at the low frequency ones, the have a very broad shape, uniform, that changes slowly with space or time. By similitude with these waves this first layer gets the name.

According to color, all the picture is mostly green (and variations). What conforms the picture is the abrupt variations of contrast on the borders. Picture by Mitya.

If we focus again on our low frequency picture, we notice that something is missing to achieve a high quality picture: details. The details such as the lines of the nerves in the leaf or perhaps a very small bird in the background have the characteristic that they are not wide and extensive, but instead are very small regions that change abruptly in very few pixels. The nerve of the leaf is also green, with the same hue that the rest of it, but on the side it gets darker because of the lighting shadow, while on the other side it gets lighter green. This information compose the high frequency layer, by similitude with the homologous sound waves: changing fast with time and space, acute and well defined. In this case the high frequency layer is coded as a grey layer (where grey means “do nothing to the low frequency layer”) that contains borders, lines and other shapes in darker or brighter tones (with the information on how to modify these parts on the low key image).

Those two layers are combined using the linear light mode. This mode is contained in the contrast group, which means that its main purpose is to modify the contrast of an underlying image. As we have our low frequency layer behind and out high frequency layer on top, the latter one will modify the contrast in the designed parts of the image to provide. Linear light works leaving the pure-grey tones of the high layer exactly the same as in the low frequency. If we have a light grey or white on the high layer, we will get a linear dodge, meaning that will brighten the luminosity of the lower layer adding the value of the upper layer. This leads to an increase in luminosity but a decrease on saturation and intensity of color. On the other side if the upper layer is dark-grey or black, it will apply the same as with the white part, but inverting the brightness after that, so we get a darker effect with colors more saturated.


That sounds difficult! How can I do it?

It is simpler that it looks to achieve the frequency separation. You don’t really need to know how it works to use it, although knowing what parts are implied in the process allows us to twitch it sometimes a little bit, allowing us to achieve some interesting effect. If you want to experiment is a whole world right in front of you, but if you want a fast and ease result, here is the recipe for success:

1- Create a new empty group.

2- Duplicate the base image twice inside the group. Rename the upper one as “High frequency” and the lower one as “Low frequency”. If you are lazy, remember that renaming them takes less than a minute and will avoid a ot of confusion when you work with more layers.

3- Turn off visibility of the upper layer and in the lower layer apply a Gaussian blur. 4 px is a good starting value, but you should choose a value that blurs all the fine details of the image but still makes all the overall image clean. Finally, turn on again the visibility of the upper layer.

Parameters for 16 bits/channel.

4- Select the high frequency layer and go Image > Apply Image. As a layer you should select the “Low frequency” one, channel in RGB, Opacity at 100%, scale with value 2 and Preserve Transparency and Mask… unchecked.

If you are working on 8 bits/channel mode (and I don’t know any reason why you should be doing this): Use Subtract as the blending mode and set the offset to 128. Invert should be unchecked.

If you are working on 16 bits/channel mode: Set the blending mode to Add and the offset to 0. Invert should be checked.

5- Change the blending mode of the upper layer to Linear Light.

If everything went well, you should now have an image that looks exactly the same as the initial, but separated into two layers, one that provides the color information and other that provides the texture information.


Why bothering doing all this?

So… I followed all the steps and I got an image that looks exactly the same as the initial one. Why should I bother to follow all this procedure?

The fact is that every photographer chooses what to do now. The separation allows you to modify those two parts of the image in different ways as needed. Do you want to correct the color of a part of the image without altering the texture? Just modify the lower layer until it fits your necessities. Do you want to get rid of some skin imperfection? Apply the healing brush or the cloning tool on the upper layer and voilá… it’s gone!

As an example, I present a picture of a portrait in the phases before and after using this technique, so you can compare the differences.

Comparison Freq. separation
She is Icxiuh, a nice Mexican chemist I worked with a few months ago. On the left, the picture has gone trough a RAW development and minor brightness and color corrections. On the right is the same picture after applying frequency separation and corrections. The skin in the forehead, cheek and the part above the lips was corrected using the cloning tool, mostly on the low freq. layer but also sometimes in the high freq. layer. To unify the color of skin afterwards the skin was softened using a gaussian blur with 8 px only in the low freq. layer. Finally, the cornea was whitened softly in that layer. I needed to work on the high freq. layer for the nails, due to their high contrast with the surroundings.

What do you think? Do you know any variation or trick related to this technique that you want to share? If so, the comments section is open to you.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s