DSLR Color Noise

[bhwolf]

Hi all,

I have seen this problem in a number of shots -- basically large scale color noise like in the attached crop.  This is overprocessed and zoomed a bit to illustrate, but having a bear of time getting rid of the noise.  I've tried ACDNR and TGVD with some success as you can see from the before/after, but it still has a smudge/smear like look to it.

This was a stack of about 40 2min exposures at f/2 (Hyperstar), with some light pollution, using a DSLR with Astronomik CLS-CCD filter.   I wonder if perhaps I'm just overexposing the images, though the histograms looked ok.  The other images I took that night don't quite suffer this issue to this degree and seem to be dealt with pretty well using the standard techniques, but those targets needed much less exposure time per sub.

Oh, frames were reasonably dithered, too, verified from star alignment showing a slight offset.

Any suggestions?

Thanks!
Brian

[pfile]

well, despite the dithering, that looks like the classic "rain noise" caused by polar misalignment drift combined with undercorrected hot pixels in the calibrated subs.

one thing you can do is use the Blink tool to watch the unregistered frames and verify that indeed there is a drift in the images.

one strategy for the hot pixels is to use CosmeticCorrection to clean up the calibrated but unregistered subs. another strategy is to do more agressive pixel rejection during integration.

rob

[bhwolf]

Thanks Rob -- good tips.  I really thought my polar alignment was good, but you were right:  I ran through the blink tool and indeed you can see some rotation.  I'll try some more aggressive rejection but otherwise will chalk it up to a lesson in alignment.

[Nocturnal]

Differential flexure already has a dither like effect so I generally don't recommend it until you have DF under control. First get DF below half a pixel for your target exposure length. Then add dithering and some kind of Sigma stacking to help reduce the effect of hot pixels.

Generally, when you see streaks like this it indicates two problems:

- dark subtraction is leaving fixed patterns behind (random noise would not result in recognizable lines)
- you have differential flexure

[pfile]

this is true, while guiding a linear drift is generally due to differential flexure; a rotational drift will be due to polar misalignment.

rob

[bhwolf]

I stretched the image and used aberration inspector to create a mosaic of the final integration, the rotation of the noise is pretty obvious.  This was about 4 hours of integration, which is a lot higher than I typically do when using the Hyperstar.  I'll need to be checking my alignment...