PixInsight Forum (historical)

PixInsight => General => Topic started by: glatiak on 2019 January 24 12:25:38

Title: Apparent Image grid
Post by: glatiak on 2019 January 24 12:25:38
Being a novice with PI, and only a couple of images massaged so far, I have noticed that after drizzle integration my stacked images have a distinct grid as though seen through a screen. I assume I am doing something wrong someplace and have made a few unsuccessful attempts to remove it but to no avail. The source is a Mallincam DS16c -- debayered in PI. Sensor is 4656x3518 pixels, bin 1. The attached image is a crop -- roughly 1/4th the total frame. Any thoughts or suggestions as to how to not have the window screen effect?

Side comment -- colour was almost non-existant, assume I underexposed the thing...

Thanks,

greg latiak
avalon observatory
Title: Re: Apparent Image grid
Post by: cliff_n on 2019 January 26 08:50:27
This is a "Me too" post.  I'm getting the same type of grid pattern during image registration (Star Alignment).  This is the same image taken at different steps.  Other images in the group (not shown here) also have a grid pattern but sometimes it is larger and sometimes it is smaller.
Title: Re: Apparent Image grid
Post by: dhb2206 on 2019 January 26 09:29:07
Are you debayering before integrating? Looks like a debayer pattern to me. Though I notice you mention debayer - is it perhaps drizzling from CFA or not as the case may be?
Title: Re: Apparent Image grid
Post by: wadeh237 on 2019 January 27 13:35:22
I see this (with my non Bayer camera) if I stretch the background too far.

I believe that it's an artifact of imperfect calibration, combined with stacking dithered frames.  In my images, it looks like some very faint residual of the fixed pattern noise (as seen in a stretched bias frame).  When you align the images, I think it's like an interference pattern (or actually an interference pattern).

In my case, I have to stretch beyond all reasonable measures before it's noticeable, so it is not an issue at all.

Note that I don't think that has anything to do with the color filter array because the scale is way too large.
Title: Re: Apparent Image grid
Post by: glatiak on 2019 January 31 10:49:03
Don't know about other folks, but I found my grid was coming out of drizzle-integration. So by reducing the scale and drop shrink I could make it completely disappear. The later appears to be the driver of the grid. Another 'user error on device' problem -- or perhaps 'de fault is always de user'.

greg
Title: Re: Apparent Image grid
Post by: dhb2206 on 2019 February 04 00:21:00
Just had a thought. You have set your raw up for Pure Raw on the Format Explorer?
Title: Re: Apparent Image grid
Post by: ApolloLaunchControl on 2019 February 11 17:16:14
And I'm getting one on a much larger scale. 3 nights of mono NB images with the ASI1600MM, drizzled, DBE, etc. and integrated with ChannelCombination. I didn't notice the pattern until I did a StarMask, and then it really popped out.
Title: Re: Apparent Image grid
Post by: pfile on 2019 February 11 17:21:48
if this happens with mono images then it is likely a calibration problem - the sky background is so weak in a narrowband image, that dark subtraction sometimes leaves negative values (which get clamped to 0) in the calibrated frames.

you can try adding a pedestal of like 100DN in ImageCalibration and the problem should resolve itself.

rob
Title: Re: Apparent Image grid
Post by: mmirot on 2019 October 30 08:09:41
I see this in my narrow band images with registration on auto/Lacnos3. I changed to Mitchell and it goes away. I would like know the cause and how to fix it before registration.

Max
Title: Re: Apparent Image grid
Post by: mmirot on 2019 October 30 09:01:18
I will try to add a pedestal. What is the proper method to do this ?

Max
Title: Re: Apparent Image grid
Post by: pfile on 2019 October 30 09:02:02
add a pedestal to the calibrated frames with the Output Pedestal section of ImageCalibration. try maybe 100 or 200. this will prevent the pixels from getting clamped at 0.

rob
Title: Re: Apparent Image grid
Post by: mmirot on 2019 October 30 10:34:15
 Do I need to do this to make the master bias and dark too?

Thanks

Max
Title: Re: Apparent Image grid
Post by: bulrichl on 2019 October 30 10:47:06
No. (I am assuming that you prepare the MasterDark by integration of the dark frames, i.e. NO calibration of the dark frames, NO calibration of the integrated dark frames).

Bernd
 
Title: Re: Apparent Image grid
Post by: mmirot on 2019 October 30 10:51:24
To be clear,  integrations of the masters without adding 100 adu etc?

Max
Title: Re: Apparent Image grid
Post by: pfile on 2019 October 30 13:08:49
only do the output pedestal when you are calibrating lights. adding pedestals to darks/bias will just make this problem worse...

rob
Title: Re: Apparent Image grid
Post by: M44DSW on 2019 October 31 13:44:42
I am seeing this as well. Not seen before, same workflow, same camera, don't think I have changed any settings, only change is PI has updated a few times since the last time I used it last season. I only get this when I Star Align, it is not apparent in the calibrated image, nor the calibration frames or the original image. I have a CMOS mono camera so ZWO ASI1600MM.

I have attached an STF stretched screenshot.

Any ideas?
Title: Re: Apparent Image grid
Post by: pfile on 2019 October 31 15:17:03
have you tried making new darks?

you can try the pedestal thing and see what happens.

rob
Title: Re: Apparent Image grid
Post by: M44DSW on 2019 October 31 15:26:01
Hi,

Ok I have worked it out, it is the algorithm being used for image registration. There has been a change in the latest update and I updated my Mac last week. So I have the interpolaion set to Auto. This should select Mitchell-Netravali Filter which it had been up until now as my scaling factor is 0.3, however it is now selecting Lanczos-3 and causing the issue. Selecting Mitchell-N manually fixes the problem and looking back at my last image I took with this camera you can see the image registration module has changed, the current one I am using is 01.04.01.0362 whereas the previous one was 01.00.05.0608. Not sure what has changed but going back to that module on another Mac fixes the issues with auto. Out of interest Nearest Neighbour also produces good images without the pattern.

Thanks
Dave
Title: Re: Apparent Image grid
Post by: pfile on 2019 October 31 16:25:40
you should still check your calibrated subs for pixels clamped at 0 - the artifact is from registration but you may just be masking it by changing the algorithm. i think lanczos-3 has been the default interpolation method for a long time now.

rob
Title: Re: Apparent Image grid
Post by: mmirot on 2019 November 04 06:02:40
Adding a 100 output pedestal at calibration does not help. Must be something else.
Title: Re: Apparent Image grid
Post by: pfile on 2019 November 04 09:38:43
well still try applying this pixelmath to one of your calibrated subs and see if you have a bunch of white pixels in the result:

iif($T==0,1,0)

you can try this on your non-pedestal calibrated image and the pedestal-calibrated image and see if there is any difference. it could be that a pedestal of 100 is not enough.

or, it could be something else as you say, but it's worth looking for clamped pixels.

rob
Title: Re: Apparent Image grid
Post by: mmirot on 2019 November 05 06:48:17
Rob,

I have used an output pedestal of 100 and the maximum of 999.  100 leaves about 70 zero pixels. 999 leaves about 25 .
Having trouble producing a cal image with none. Suggestions ?


Max
Title: Re: Apparent Image grid
Post by: pfile on 2019 November 05 07:56:46
how about an image with no pedestal applied?

those are small numbers of 0-valued pixels, i think for the grid to show up as a result of clamped pixels you'd need a lot more of them than what you have. there could be some other cause for your problem...

rob