One of the more advanced and challenging workflows is to remove the stars from astronomical images. Why would you do this? Stars are part of the deep sky image.
An astro photo, especially nebula photos, takes on a whole new look when the stars are removed. You are able to see more detail and more structure.
Although you expect to see stars in any astronomical photo, depending on the target, the brightness of the stars might overwhelm the subtleties of the object. With PixInsight, we can reduce the brightness of stars or completely eliminate the stars. Although morphological transformation is used in star reduction, I follow a different process for star removal.
TIP: This is one approach I use for star removal. Another option is to leverage StarNet. For reflecting telescopes, StarNet does not perform as well as it does for refractors. However, this Star Removal technique with StarNet works really well.
In order for the star removal process to work, we need a very good star mask. Follow the guide on creating a star mask.
For really large, bright stars, it is often best to remove them from the star mask so they remain in the final image. These are extremely difficult to remove and blend with the background.
Once the star mask is created, we need to apply it to the image so everything, except the stars, are protected.
At this point, we need to zoom and cycle the mask on and off with Control+K. We need to make sure the star completely covers the stars, all the way to the edge. If not, the star removal process will leave a halo.
If the size of the stars in the star mask is not large enough, run Morphological Transformation in dilation mode.
The first part is to use the morphological transformation process to eroded (shrink) our stars. In this case.
Within morphological transformation, there are a few options to be aware of:
- Operator: For star removal, we want to use Erosion
- Interlacing: Adding interlacing creates a noisier result, which might seem odd. But when I do star removal early on, I have yet to do noise reduction. To blend better, I want a noisier result.
- Iterations: Number of times to run through the process. Using a iteration of 1 allows us to see the impact with each step.
- Amount: Determines how much blending there is with the new values vs the old. For star removal, we want to do a complete removal, which means a value of 1.
- Structuring Element: For my imaging rig, using a size 7 as a circle works best
I continue to rerun this process (3 or 4 times) until the stars are removed.
Multiscale Median Transform
By using Morphological Transformation first, it becomes easier for multiscale median transform to hide the remaining parts of the star. I find that using Multiscale Median Transform does a better job at completely removing stars when compared to Morphological Transformation. Multiscale Median Transform creates a smoother image while Morphological Transformation starts to create dark spots if it is applied to many times in order to darken the very bright stars.
If I only used Morphological Transformation, I would then have to find a way to fill in the dark spots.
To use Multiscale Median Transform, it is a good idea to use the image preview window.
Increase the number of layers to 8.
Start at layer one and disable. Keep disabling the next layer until the stars stop dimming.
For my setup, I usually stop at 7 layers as the 8th does not appear to have an impact.
Apply this to the image multiple times. Each application results in the stars getting dimmer. Do this until the changes between attempts is minimal.
I am noticing that my background is littered with many tiny, tiny stars. To remove those, I will create a new star mask to capture these tiny stars and rerun this star removal process.