Bolton Astronomical Society

Grainy images
Page 1 of 1

Author:  bchamberlain [ Mon May 11, 2015 1:41 pm ]
Post subject:  Grainy images

I'm wondering whether either of these settings affect the amount of grain in my images processed in IRIS.
1: What should the normalization value be for 14 bit RAW files. (can't find it on the WEB)
2: Polynomial v Local estimator. Which is best and does it influence grain?
Thanks for any advice re grain. Bill C

Author:  rwilkinson [ Mon May 11, 2015 2:54 pm ]
Post subject:  Re: Grainy images

In his tutorial ( Christian Buil (the author of IRIS) recommended a normalisation value of 5000 for 12-bit data (i.e. 4096 levels), but perhaps it should be different for 14-bits (16384 levels)?
Buil suggests running the STAT command to check the maximum pixel intensity of an image (this must be kept below 32767 to avoid saturation). So check your calibrated images (i.e. after offset, dark & flat correction, but before CFA conversion) and if your "Maxi:" value is anywhere near this limit, try a different normalisation value?
If you do hit this limit, you'll saturate the brightest parts of your image (i.e. stars), and so compress the overall signal-to-noise ratio.

I can't see how any gradient reduction process could create small-scale "grainy" artefacts in the image background (unless maybe you used a very small value of computation area in the local estimator?).

But remember that your images start as 14-bit RAWs, and IRIS processing operates in a 16-bit enviroment, so make sure that you've not inadvertently truncated down to 8-bit resolution (i.e. only 256 levels in each colour-plane) when transporting your data between the other post-processing programs which you use.
The reduction to 8-bit JPEG should only be done at the very last stage (if you need to reduce the file-size to upload to the web-site).

Author:  DRatledge [ Sun May 17, 2015 12:58 pm ]
Post subject:  Re: Grainy images

I assume you mean normalisation value for making flats? As Ross says the default of 5000 is set for 12 bit images. For 14-bit images I use 8000 - being the half value. The exception is narrowband hydrogen alpha images - I use 800 for that. It shouldn't noticeably effect graininess.

Light pollution removal - every image is different - sometimes Polynomial works best - sometimes Local Estimator work best. I always try both and use a mask.

Light pollution removal subtracts a smooth light pollution pattern. But the light pollution we have recorded in our image is not smooth - at each pixel there is a random variation - this is called noise. It is why our images from Lancashire are much noisier than those from a dark sky. It is a cross we have to bear and I keep asking myself - why have I not moved to Norfolk?

Page 1 of 1 All times are UTC
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group