RGB noise - help

[dougj7]

Hi,

My images are suffering from red green and blue dots. I am using a Atik 320 color ccd camera. Images captured and debayered in nebulosity 2. Star alignment and image integration in PixInsight 1.7. Can someone tell me how to get rid of the rgb dots and what are they caused by? Thanks for any input.

-Doug

[dougj7]

here's another image at 1:1 scale. mostly red and blue with very few green dots.

[Carlos Milovic]

I think that you have hot pixels, not removed by the darks. I would try to calibrate the data before debayerization, and eliminate those pixels there.

[dougj7]

Thanks Carlos,

I was hoping there was a feature in PixInsight similar to deepskystacker "detect and clean hot pixels" feature. 

[Carlos Milovic]

Well, you may try standard noise reduction techniques but definitely it will be better if you remove them from the image before you debayer them. Since the DefectMap process don't work too well with bayered images (today I'm starting a 3 weeks vacation, so I might rewrite it to handle them) I would encourage you to try the CosmeticCorrection scrip. You may use the same master darks to create the needed "defect map".

If there is no coming back, to recalibrate the data, then, as I said, try ACDNR or wavelets. You may help a lot the process is you build a suitable mask. In this case, I think that if you work over the CIE c channel, you may be able to isolate those noisy pixels. Then, use morphological filters to improve the mask.

[harist]

I get these hot pixels as well from time to time and I don't use darks with the SXVR-25C. I've found that median combination in ImageIntegration removes them (at the expense of lower SNR than average combination though).

Tasos

[Nocturnal]

Remove hot and cold pixels from FITS files with my FixFITS utility for windows:

http://www.tungstentech.com/Software/FixFITS/tabid/78/Default.aspx

It also finds and fixes bad columns, even subtle ones. I always pre-process my QHY8 lights with this utility before stacking.

[gvanhau]

Thanks Sander for your contribution.
Could you explain your algorithm? may be we could make a script in PI. Or is this similar to cosmetic correction script?

Also related to this topic is this one: http://pixinsight.com/forum/index.php?topic=3238.0

Regards
Geert

[Nocturnal]

Hi Geert,

I've thought about converting the code to a module but frankly that's like using a bazooka to hunt rabbits. Writing modules is hard work and it doesn't pay. The utility is available for free already so I think that's good enough.

The hot/cold pixel removal tool simply loads a dark frame and finds the outliers. These same locations are then averaged in the light frames. In OSC mode (default) it takes pixel values from 2 pixels away rather than immediate neighbors.

I think this is a better approach than trying to use wavelets or some other mechanism to find hot pixels in light frames. Hot/cold pixel detection in DSS also doesn't work very well in my case because I don't use dark frames for calibration. Actually I don't know how DSS does hot pixel detection,I think through slope detection rather than looking at dark frames. As long as the dark frame matches the light frame reasonably well this work very well. Or 'perfectly' as Juan would say. This depends on the camera of course but I took a few dark frames 2 years ago and they're still effective even over wide temperature ranges. I sometimes adjust the sigma to get more or fewer hot pixels.