›

Background Equalizer

DAVID JENKINS
in PixInsight
After watching Adam Block's Fundamentals series of videos on Pixel Math, I was inspired to tackle a nagging problem that has been bothering me with my astrophotography images. When I finish processing, I tend to zoom way into my images to examine the quality of the sky background (also known as pixel peeping). I'm usually disappointed to find bits of noise, and sometimes artifacts, that never went away. For me, trying to do any further denoising at this level usually results in creating a bigger mess. However, using Pixel Math, I think I have come up with a simple solution to equalize the sky background without impacting the objects in my image. Here goes: I went to Adam Block's image gallery to download a couple of his wonderful images: Nebulae -> Cederblad 51 and Galaxies -> M51. In PixInsight, I zoomed into the corners of these two images and discovered that there existed this same dreaded noise. This is my new method: I examine the image to find the faintest possible detail that I want to protect. I then measure the brightness of that detail to establish a baseline. In these two images, this appears to be around 0.07 to 0.08. Therefore, to be safe, anything dimmer than 0.06 contains the pesky noise. Now, using Pixel Math, I enter a single RGB/K expression as follows to equalize the background: iif($T<0.06,0.06,$T). Of course, the threshold of 0.06 could be adjusted up or down, depending on the image. However, in these two "51" images, I found that this seemed to work well. The end result is a wonderfully smooth background sky with what I feel is a nice level of contrast. Now, I sleep better at night. :)  What do you think?

Comments

  • Roger Smith
    Hi David,
    I just used the expression with my own values on my image after DBE (several times) and the background is definitely uniform everywhere. But to me it is too perfect. The perfect background accentuates the noise in the galaxy. 

    I did improve the too perfect situation (in my opinion)  by applying the NoiseGenerator process. I had a star and galaxy (Game) mask which I applied to allow NG to be applied only to the perfect background. I found the Gaussian to give the best look, followed by SCNR and a little boost to the blue. A very mild Convolution also helps it look more real.

    But I still have my noisy galaxy! Thanks for sharing your idea.
    Roger
  • DAVID JENKINS
    Hi Roger. Thank you for trying this out. Yes, if your main subject is noisy, it may be too much of a stark contrast. I find it works well as the final step in my processing to eliminate the residual background noise. You're right - this would not be a good idea on a generally noisy image.
Sign In or Register to comment.