De-Bayering or Dithering problem?

[fwm891]

I've been imaging the rosette (6x5 min subs) and I'm getting a linear pattern to the background.

Capture is via Maxim and I'm employing a dither of < 2pixels between exposures. No filteres have been used and the DSLR is unmodified.

Images have not been de-bayered, simply registered and stacked in PI then processed in PS-CS3.

My thoughts are that there is a resonance between the dithering and the rows/columns of pixels on the CCD. But it may be that I need to de-bayer the images?

Can anyone throw a light on this please

[lensman57]

I've been imaging the rosette (6x5 min subs) and I'm getting a linear pattern to the background.

Capture is via Maxim and I'm employing a dither of < 2pixels between exposures. No filteres have been used and the DSLR is unmodified.

Images have not been de-bayered, simply registered and stacked in PI then processed in PS-CS3.

My thoughts are that there is a resonance between the dithering and the rows/columns of pixels on the CCD. But it may be that I need to de-bayer the images?

Can anyone throw a light on this please

Hi,

If the images have not been debayered then you would not have colour, those look like vertical banding noise to me and oyu should be able to them in PI .

Regards,

A.G

[frugal]

That pattern looks just like the Bias noise from my Canon DSLR. Have a look at this excellent article by IanL

http://www.blackwaterskies.co.uk/2013/09/pixinsight-dslr-workflow-part-1-bias.html

[Coco]

That's a great link!

[Uranium235]

Agree that it looks like bias noise. If the vertical banding is still present after calibration, noels actions can deal with it (vertical or horizontal banding removal).

[opticalpath]

I don't think we answered the part of the question that Francis asked about dithering and why that didn't eliminate the pattern noise.

How wide are the noise bars in an un-dithered summed image? If more than 2 pixels, I suppose you would really want to dither more than that to get the full benefit.

I would also inspect the Maxim guide log just to check if dithering was working as you expected: i.e. do you see big enough displacements in RA at the start of every exposure, and do they vary enough, keeping the x-position of the guide star changing and not settling back to the same positions repeatedly?

Adrian

[Coco]

Are we all still using BIAS frames, am sure on here somewhere I read it was pointless with Canon CMOS chips as it subtracted the bias from the lights before finishing the recorded image ?

[IanL]

I still use bias frames with the DSLR.  It is true that there is no need to the bias from a DSLR / CMOS sensor as correlated double sampling eliminates the need for a bias to be added prior to exposure. For many models there is still a fixed pattern that you can remove by following the bias calibration process. Easy way to check is to make a couple of dozen bias shots (i.e. covered camera, shortest exposure possible), stack them, stretch the image and take a look to see if a pattern is emerging.

I can't see any particular reason not to use them as it is fairly trivial to make a large master bias that you can use for all images thereafter. Just take care that you are not ending up with truncated negative values after subtracting the bias (so in PI you would check to make sure you don't have a much larger number of zero value pixels after bias subtraction than you did before; if yes then you should add an output pedestal at the bias calibration stage).

With regard to the problem in the image above, I assume it is a 383L+ image, so I'd be wary of drawing any direct comparisons with a DSLR sensor.  That said, it may be that the dithering distance is just too small and some kind of noise component is resonating? What would be helpful is if you could post your ImageIntegration settings (or BatchPreProcessing settings if that is what you used).  Also take a look at the high and low rejection images if you have them - could be that your pixel rejection parameters are not restrictive enough and failing to eliminate the pattern.

[Coco]

The image above is a DSLR otherwise I would not have made the comment.

I am reasonably confident with PI as far as Mono is concerned but the DSLR job has got me foxed for the time being anyway..

Think its best to convert the CR2's into .FITs then start calibrating & registering... just wished it was as easy as the 428 OSC i used to have..

[Thalestris24]

Hi

You didn't mention darks or flats? They are essential! Also, to get the best advantage from dithering I think you really need at least 10 subs. The more, the better. You could try stacking in DSS (the beta version for Canon CR2 files). It's free and quite easy and quick to use.

Hth

Louise

[IanL]

Good point re the number of subs as well.  What pixel rejection algorithm do you have selected in the PI integration or BPP process?  For six subs you ought to be using percentile clipping to get the best results. The other methods need more subs to work well (or at all in some cases), but percentile usually does a good job on small sets of subs; play with the percentile low and percentile high values if the defaults don't yield the right result.