Processing total eclipse images

[RickS]

I have a set of 171 images taken at the recent solar eclipse in northern Australia that I have been attempting to process in PixInsight.  They were taken during the period of totality with a DSLR and 200mm lens on a fixed tripod at a range of exposures (using in-camera bracketing).

So far the best result (see attached image) was produced by using the FFTRegistration script to register and sum small batches of images.  These summed images were processed with the LarsonSeknina process to bring out the details of the corona and then registered manually in Photoshop (so that the coronal features coincided) and combined into a single image.

This was a fairly painful manual and experimental process, so I didn't actually use all of the subs.  I'm planning to go back now and process all of the subs to see if I can get a better result and I'm looking for helpful suggestions on how to do the registration more accurately.

The problem is that I need to register all of the subs against the corona, not the moon, since they are moving relative to each other.  FFTRegistration does a great job of registering the subs against the moon but that's not what I want.

One idea I have is to measure the displacement between my first and last image and then use the image timestamps to interpolate delta x,y values for each image and use a script to do a translation on each image.  If anybody has a better suggestion I would love to hear it!

Thanks,
Rick.

[Carlos Milovic]

If there are stars in the frames, you may use them as reference, with DynamicAligment. I did that with the solar eclipse from easter island.
The most tricky part is to do the HDR processing... I'll comment on that subject later.

[RickS]

No stars I'm afraid, Carlos.  That would have been too easy.  I expected to see some but I think they were wiped out by high cloud.

I'd be interested to hear about how you did the HDR processing when you have time to explain.

Thanks,
Rick.

[RickS]

One idea I have is to measure the displacement between my first and last image and then use the image timestamps to interpolate delta x,y values for each image and use a script to do a translation on each image.

That didn't work too badly.  It's a shame that the EXIF data doesn't have date/time to better resolution than a second.  A deeper image using the rest of the data is attached.  I'm sure I can get some more out of it with additional processing!

[Philip de Louraille]

Nice!

[RickS]

Thanks, Philip!  I'm going to try a few more experiments and see if I can get some more detail out of it.

[RickS]

Here's the final result.  I'm quite pleased with it, considering that conditions were far from ideal with high cloud and the sun low in the sky.

Cheers,
Rick.

[FunTomas]

Hi RickS. How did you register your images? I have set of images from this solar eclipse, but have no idea how to register these images if there are saturated areas. FFT registration totally failed.

[RickS]

Hi RickS. How did you register your images? I have set of images from this solar eclipse, but have no idea how to register these images if there are saturated areas. FFT registration totally failed.

G'day FunTomas,

FFTRegistration did work for me.  I didn't have to do anything special.  It's not ideal because it appears to register on the moon and not the coronal features.

I ended up using another method that should work for you: take your first and last images and run a LarsonSekanina process on both to bring out the corona details.  Register these two images manually and calculate the number of pixels you had to move the second image in X and Y directions (I did this in Photoshop using a difference layer so I could see when the images matched best).  Divide these numbers by the time difference between when the two images were captured (using EXIF time stamps).  Using this data and the EXIF timestamps in your images calculate the delta X and Y that you'll need to register each one.  Transform each of the images using PixelMath, and you're done!

Cheers,
Rick.

[FunTomas]

Hm, good idea (to use Larson-Sekanina). I will try it today, but at the evening. Thank you.