Thursday, August 31, 2017

Eclipse Corona Image

Here is my main image from the solar eclipse:
(click on image for full resolution)

This was my biggest challeng: to process the 3 bracketed image sets that captured the corona of the sun. The corona has a very high dynamic range. My first tries to use any HRD algorithms (in Lightroom, Nik HDR Efex...) did not go very far.

I found a couple of useful tutorials on the web:
They all employ a similar workflow:
  1. Align all images precisely
  2. Create a composite image of the sum of all individual images
  3. Run a radial blur filter or a Larsen-Sekanina filter of the composite. This will create an image with all the detail but low contrast (mostly grey or black)
  4. Multiply the sum image with the detailed, low-contrast image
  5. Make final adjustments (stretch, curves...)
Fitswork has an interesting approach: pick two images, overlay by subtracting one from the other, this will make it easy to align them. It does work. ... but I found it too cumbersome. I was looking around for another solution and found the FFTRegistration (Fast-Fourier-Transformation) script in Pixinsight. This is often used to align comet images.

Enter a reference image, add all the images, I wanted to store the registered images, so I checked this and click "OK". ... takes a while and this was the result:

I took this image into FitsWorks and selected the Larsen-Sekanina Filter:

There are only two settings:

Rotation and Radius. The tutorial recommended to start with Radius=2.0 and Rotation=1.31. I chose those settings and got this result:

It has a lot of detail - and almost not other information (low contrast, wrong colors...)

But multiplying both images gives this result:

Doing a simple stretch:

And some curves, saturation and color adjustments:

And then some final tweeks in Lightroom (devignetting, cropping...)