Global Star Alignment (Deep-sky) [Siril 1.0]

Global Star Alignment (Deep-sky)

This is a more powerful and accurate algorithm (but also slower) to align deep-sky images.

The global matching is based on triangle similarity method for automatically identify common stars in each image. A linear transformation is then computed using RANSAC algorithm,to further reject outliers and determine the projection matrix.

Note

A new sequence is created with the prefix of your choice (r_ by default).

Screen

List mode Interactive mode

1⚓
Layer on which the registration is done.⚓
It can only be done on one layer at a time.
Tip
The (*) sign appearing after the layer's name means that registration data is already available for this layer.
Note
When processing images, registration data is taken from the default layer if available (for RGB images: Green, else fallback to Blue then Red).
2 Prefix⚓
By default, prefix 'r_' is preprended to the sequence name. You can specify another prefix there.
3 Interpolation algorithm⚓
Type of algorithm used for interpolation⚓
Choose algorithm:
Nearest Neighbor
Bi-linear
Pixel Area Relation (recommended)
Bicubic
Lanczos-4
4 Match stars in selection⚓
Match stars in selection Check this box if you want to perform the Global Star Alignment algorithm within the selected area in the reference image. If no selection are done, this option is ignored.
Warning
Make sure to draw a selection with a sufficient number of stars.
5 Save translation in seq file⚓
Note
Activating this is somewhat similar to using one-star registration, except that a larger number of stars is used to compute the shifts.
Advice
When this option is enabled, it is recommended to use Shift in the Transformation box.
Tip
This is useful for image sequences with very small rotation angle.
Warning
No new sequence is built, registration data is stored in the original sequence,with subpixel accuracy. However, pixel shifts are used for stacking.
6 Transformation algorithm⚓
This list specifies the selected alorithm to compute the transformation between the images and the reference image. You can choose between :
- Shift
- Similarity
- Affine
- Homography (default)
Tip
By default, the transformation is a homography, which admits a certain number of distortions between the 2 images. It is a “flexible” model which adapts very well to the general case.
But for sequences of images in which one would expect little or no distortion, this projection has too many degrees of freedom and one may wish to test more rigid projections, which will result in less interpolation. To lower the number of allowed distortions, you can choose a translation (1 translation along each axis), a similarity (2 translations, 1 rotation and 1 scale) or an affinity (the same with shearing in addition).
Back to screen
7 Minimum pairs⚓
This input specifies the minimum number of star pairs to validate that the registration algoritm has suceeded.
Advice
While for the general case, the default value of 10 should work, you may have to lower this value if registration fails, i.e. if there are not that many stars on your images.
Beware not to lower it down too much as it may accept false registration results.
The absolute minimum of pairs to be matched also depends on the Transformation algorithm selected. For instance, a Homography being a 8-degree-of-freedom transform, a minimum number of 4 star pairs is required to define it.
Back to screen
8 Adjust star detection⚓
If you need to increase the number of stars detected, pressing this button will open the Dynamic PSF dialog.
Dynamic PSF (for registration settings)
Back to screen
9 Go register⚓
It will register images in the loaded sequence using the selected method and reference layer.
Advice
Loaded in the 'Image sequences' tab, current sequence name is at the bottom of the window too.
Back to screen
10 Simplified Drizzle x2⚓
When ticked, upscales the images by a factor of two
Advice
With this registration method, a new sequence is created. If this option is selected, the new images will be four times a large as original images, which may take a lot of space and require more time to compute.
Back to screen