Some seriously good tips in here. I have long wondered how to get better colors in the sky when processing blue skies. Between that and @s7habo’s tip about maksing, I may take another crack at a photo that gave me a lot of grief, both because of the issues with making an accurate mask, and the poor color in the sky.
Here’s the shot. You can see artifacts where I tried to bring more blue out of the sky, and the masking around the trees is shoddy. I’m embarrassed to say how much time I spent trying to get this photo right.
One question: This is the second time in 2 days that I’ve seen someone using the PCA Mixer from G’Mic. Where do you guys learn to use these things? I tried it out and mashed a bunch of buttons for 15 minutes, couldn’t figure out what it was doing or how to control it, and thus will probably never use it. If I had some guidance or documentation, though…
Basically the mixer considers 3 PCA components. The primary contains most of variation of the image, followed by the secondary and finally the tertiary. I.e., you could remove the tertiary and still see most of the original image. My first impression is that factor is the weight of the components, shift is the intensity and twist is the hue shift.
Since GIMP 2.10 you can set blend modes for the layer groups .
This can be very useful.
An example for high pass sharpening (How to arrange the layers and which blend modes are used can be seen in the picture below):
Now, if you move the slider for Gaussian blur, you can observe the sharpening radius directly on the canvas. This way you can precisely adjust the radius (size of blur) and amount of sharpening (opacity of Group layer) before applying it.
If the sharpening is too strong, instead of linear light you can also use grain merge or overlay blend mode for the group layer.
This is a great tip. Recently, been selectively sharpening my photos and applying different sharnesses to different parts of them. Is your procedure basically the same thing that “unsharp mask” does, but the long way around?
Not really. Although the final effect is similar, the big difference lies in the blend modes used for both approaches and the corresponding consequence for further processing.
Difference in composition can be seen here:
As the advantage of Unsharp masking is its ability to control the contrast and width of the edge and which area outside the edge should not be affected, the strength of the high pass filter lies in the possibility of using different blend modes with corresponding influence on the end result.
Here are a few examples (left is the result of Unsharp mask for comparison and right are different blend modes for High pass filter).
So, each of these blend modes affects the sharpening in a different way. In my experience, depending on the image and which blend modes you use, I can often achieve better results with high pass sharpening than with Unsharf mask.
However, in the above example, the blur layer that was used to create high pass filters is preserved, and can be used for other purposes e.g. frequency separation.
Here is another example of using one color model component , this time LCH C (ab) to brighten up the image without affecting highlights and shadows too much and with some ability to control contrasts.
In this picture I want to brighten up green trees and the house without affecting white benches and texture of the cloudy sky too much:
So, we duplicate the layer and extract the LCH C(ab) component (Colours – Components – Extract Component - LCH C (ab) ):
When zooming in we see that the result is noisy. We will use Selective GussianBlur to remove the noise (Filter - Blur – Selective Gaussian Blur):
Blur radius 16,00 and Max. Delta 0,040 works fine:
For this layer we choose Hardlight as blend mode (from drop down menu in layer window choose Hard light):
We make a new layer out of the result (right click on top layer and choose “New from Visible”):
We delete Hard Light layer and rename Visible to Screen:
Now we change blend mode of “Screen” layer to Screen:
As you can see, the photo is already a little brighter. Now we will use curves to influence the brightness and contrast even more.
Open curves tool (Colours - Curves) and move upper point to center of graph as shown in the image below:
The photo is now much brighter and we have starting point for further adjustments.
We can increase the local brightness of the image even more by drawing an arc on the right side of the curve:
If we want more contrast, we carefully drag black dot very little to the left:
C is the A+B Channel? I thought that was LCH(Color). Chroma is a extrapolation of A, and B, but not a copy a transfer of A and B Channel. Am I thinking this wrong?
No. L* is the same. C*h is a polar projection of a*b*. Maybe that is why I found the naming of LCH C (ab) confusing. Moreover it is possible to get C*h from L*u*v*. Following GIMP’s naming convention, I guess C* would then be called LCH C (uv). Does that make sense?
Skin retouching with Gaussian blur in combination with HSV saturation blend mode.
Duplicate original layer, add HSV saturation blend mode to that layer, select skin area you want to clean and apply Gaussian blur
(I’ve hidden skin selection here so you can see the result better.):
Fine details are preserved and can be enhanced by sharpening the area a little bit.
Bigger skin blemishes have to be treated differently and the method is not as precise as frequency or wavelets separation, but in most cases you can get very good results quickly!
Yes, as in the picture below from the tutorial. If there are some hairs in the selected area, wavelet decompose makes sure that they are preserved when you work precisely.