Stacked colour fits

Post a reply


This question is a means of preventing automated form submissions by spambots.
Smilies
:D :) ;) :( :o :shock: :? 8-) :lol: :x :P :oops: :cry: :evil: :twisted: :roll: :!: :?: :idea: :arrow: :| :mrgreen: :geek: :ugeek:

BBCode is ON
[img] is ON
[flash] is ON
[url] is ON
Smilies are ON

Topic review
   

Expand view Topic review: Stacked colour fits

Re: Stacked colour fits

by han.k » 20 Mar 2019, 21:05

Hello Serg,

The math is nicely described in the link. Maybe useful for the future. I do something similar for drizzle but with a different and mathematical a simpler approach. I split the pixels in 10x10 so 100 sub-pixels. Then by rounding of each sub-pixel position you will find the correct destination. ( I just see it has a bug in colour mode due some other changes)

I'm aware you could calculate on sub-pixel accuracy, but that will slow down the stacking process with maybe a factor 4. Each pixel value should be distributed over 4 or 5 pixels in ratio.

Having the images 80 degrees rotated like your sample is a very rare case. In most cases the images are from one series with more or less the same angle. If you stack 20 or so images the problem will also be less since the statics will help you to correct the error.

Regards, Han

Re: Stacked colour fits

by Serg1 » 19 Mar 2019, 17:38

Hello Han.
I changed the algorithm for filtering white and black dots by adding the Average method.
https://yadi.sk/d/L2tXDFoChPvGzw

Re: Stacked colour fits

by Serg1 » 03 Jan 2019, 11:50

Hello Han.
I've tried a new method. He, too, gives swathe of dark points. I think the problem is not demosaic, but rounding pixel values when rotating images. See the link for Precision rotation of the bitmap to an arbitrary angle.
https://habr.com/post/160401/

Re: Stacked colour fits

by han.k » 01 Jan 2019, 22:18

They demosaic method "Malvar, He, Cutler" was not working for stacking. This has been fixed in v0.9.152.

I tried some other demosaic methods. They where all worse for astro images. The method "Malvar, He, Cutler" gives the lowest HFD values.

Han

Re: Stacked colour fits

by han.k » 31 Dec 2018, 09:04

To improve bright star colour, I'm working and testing new demosaic routines. In a few days I hope to have results

Re: Stacked colour fits

by han.k » 30 Dec 2018, 16:07

Hello Serg,

When the images are rotated and stacked could have places where pixels are missing. These missing position tends to be dotted lines.

Theoretical it is possible to calculate in fractions of pixels but this becomes pretty CPU intensive. If you have enough images maybe 10, all pixel positions are filled since images are all a little shifted. The program keeps record for each pixel how many times a value is added. If you have 10 images stacked, some pixels have maybe 8 values added, others have, 9, 10 or 11 values added. By keeping a record of the number of pixel added, the program can normalize the sum by dividing the amount of pixels added. So if the stacking pixel is the sum of 10 images it is dived by 10. The sum of 10 pixel values/10 is equals to sum of 9 pixel values/9. This only works if the exposure time and background value are similar. This doesn't work if you mix 10 and 60 seconds exposures. So just keep one exposure time.

I'm still not happy about the bright stars after de-mosaic. Next clear sky moment, I will connect my old QHY8 and see what results I get with ASTAP. PixelInsight is using VNG de-mosaic procedure for the same reason. It will be not easy to implement.

Han

Re: Stacked colour fits

by Serg1 » 30 Dec 2018, 14:12

Hello Han.
A new method of de-Bayer leaves the same artifacts.
Attachments
2018-12-30_17-10-00.png
2018-12-30_17-10-00.png (31.96 KiB) Viewed 26253 times

Re: Stacked colour fits

by Serg1 » 30 Dec 2018, 05:58

Hello Han.
Repeat the artifacts on the stack 4*60 has failed. This is odd. But they exist when adding 4 * 10 and 4*60. Similar artifacts exist on other stacks. I noticed that they appear if there is a small rotation of the image. When the rounding error accumulates, it is rounding up or down. Accordingly, bright and dark spots across the whole frame from edge to edge.
Attachments
2018-12-30_08-44-28.png
2018-12-30_08-44-28.png (596.99 KiB) Viewed 26257 times

Re: Stacked colour fits

by han.k » 29 Dec 2018, 20:54

Hello Serg,

I only see it occurring when I stack the 10 sec and 60 sec images together. It doesn't happen if you stack the only the 10 sec series or the 60 sec image series. I assume the internal "black spot filter" is trying to eliminate some black spots which could occur if you combine a small amount of images. By combining two different exposure times it is confused.

As long it doesn't happen with a single series of the same exposure, I would not be worried about it. Can you confirm that?

Please note note also that the background value changing quickly for each image due to twilight. That will make it difficult to remove any satellite track since it can't calculate accurate an correction. Best is to have steady conditions and same exposure time.

Han

Re: Stacked colour fits

by Serg1 » 29 Dec 2018, 18:52

4х60. Lower left corner. It's not a satellite track.
Attachments
2018-12-29_21-51-15.png
2018-12-29_21-51-15.png (80.24 KiB) Viewed 26261 times

Top