Star Trails - The Intentional Kind

[Simon Hicks]

Hi all,

Whilst on holiday I took a couple of sets of star trail images. I'm now trying to get some nice star trail images out of them. They are a set of maybe 25 x 4 minute exposures with an 18mm camera lens with a cottage and some trees in the foreground.

The question is really 'is there a way to combine them in PI?' I can't do a straight stack of the images because this enhances the S/N of the background  , but washes out the star trails. 

I've tried the excellent little program at www.startrails.de. Firstly this only accepts 8bit images . But if I convert all my raw images to jpeg and put them into this program then I end up with quite noisy images at the end. I guess that it detects the differences in each image (i.e. the star trails in different positions) and makes sure they maintain their 'full' intensity (i.e. they don't get diluted by subsequent additional images)....but I guess that this means that all the noise in each image also gets maintained at its full intensity. Noise is a difference just like a star trail in a different position.

So has anyone got any good ideas as to how to get all the benefits of normal stacking of the background and foreground, whilst also not washing out the actual star trails by stacking?

Cheers
         Simon

[Simon Hicks]

I wonder if there is a way to do two seperate stacks and combine them. The first stack would be a normal stack with sigma clipping etc to get a nice clean background and foreground....whilst largely washing out the star trails.

The second stack would be a 'difference maintaining' stack. This would take the difference between one image and the next...the difference being the star trails and the noise. This second image would have nice star trails, loads of noise and not much background / foreground.

The trick would be to select a mask for the second stacked image that only selects the star trails and not the rest of the noise / background / foreground. Star trails are typically long thin lines, slightly curved, with roughly equal intensity along their length. Is there a way to create a mask that will just select these features?

[Enzo De Bernardini]

Simo, this script may help:

http://pixinsight.com/forum/index.php?topic=2122.0

Greetings,

Enzo.

[Simon Hicks]

Hi Enzo,

Thanks for the link to the thread that I had read before my holiday...and completely forgotten during my holiday    It was a good holiday  8)

Anyway, your script does a great job of combining the raw images  but unfortunately it has the same problem as the StarTrails software, i.e. when it does the MAX operation it ensures that it always adds and maintains the brightest noise spikes from all the frames. So I get the star trails but also specially selected highlights of the worst noise from all 25 frames.

On the other hand if I do a normal stack with sigma clipping then I get a fantastic foreground....all the noise has gone...but so have the star trails.

I'm wondering if there is a way to get the best of both worlds?

[Enzo De Bernardini]

Hi Simon,

Maybe you can calibrate the light frames (master dark, master bias) before integrating and/or use a defect map to remove hot pixels.

Hope this helps,

Regards,

Enzo.

[Carlos Milovic]

Hi Simon
I would generate the "master background" image, and store it somewhere. Then, using PixelMath and ImageContainer, substract the background to all the images (don't rescale). This way, you'll create a "new signal only" set of images. On one of them, try GREYCStoration or ACDNR to reduce the noise. Once you have the right parameters, apply them to the whole set, using another Container. Hopefully this will be enough to generate a new "trails only" image, integrating with the maximum operator. To get the final image, just add the trails to the background (with PixelMath, rescaling this time).

Another alternative to the noise reduction could be a morphological filter. Use the Closing operator, with a small circular structure element.

[Simon Hicks]

Enzo, Carlos,

Thanks for the suggestions....I will try them tomorrow when I have some time and report back.

Cheers
          Simon