Hello, would it be possible to implement the following for Siril: when registering images, instead of transforming and saving the image as a new file, just save dx, dy, scalex, scaley, and rotation, or maybe the equivalent matrix transform, whatever is easier. Then, you’d calculate the transformed image when stacking. This way, you’d avoid writing large amounts of data to the hard drives / SSD, which is a significant slowdown, especially for hard drives which are usually necessary for processing large number of raw photos. Thoughts?
It is funny you ask, it has already been implemented for the 1.2 version. This also allows reframing of the sequence to happen.
Welcome to the community 
Yes, but for now the production of images is always needed for the stacking.
Maybe someday one could stack on the fly, but not yet.
Oh yes I didn’t read well. The files that would normally be excluded from stacking won’t need to be written, like those with bad stars or clouds.
1 Like
What is the reason that on the fly stacking is not possible, maybe it’s something I can contribute to
it’s very slow for several reasons, the biggest being that in stacking we read parts of images at different times, if we need to transform them before, we need to read entire images, transform them and extract the part we need, for each image and each part. It would also take more memory and require everything to be recomputed if stacking parameters are changed.
And as we said here: