PixInsight Forum (historical)
PixInsight => General => Off-topic => Topic started by: Jonathan49100 on 2018 February 12 11:17:09
-
Hello everyone.
I am a French astrophotography enthusiast who follows your great forum for 1 year, a lot of response here compared to what can be found in France, so thank you for all this help!
Here I worked with a DSLR (T3i), a 250/1000 and Azeq6 for 4 years and I encounter a lot of problem for 1 year on my treatment.
I do not think I've changed my way of working so I'm starting to have a lot of help.
Some examples:
M81 / 82 with 40.2 hours of integration (483X300s), pretraitement ok and stacking with Linear fit (some gradient on the brutes) and here is what it leaves me ...
I've tried all kinds of Rejection% and windsorized Local Normalization or ABE before the stack but nothing helps, I end up with a horrible picture.
Do not take into account the tapes, I had a dithering problem ...
(https://i.imgur.com/pB3RqtYh.jpg)
Another example with M101 and 14Hours:
(https://i.imgur.com/ay9g6Rsh.jpg)
After DBE:
(https://i.imgur.com/pi6SrnKh.jpg)
still a lot of chromatic noise ...
Each time it is a huge fight to remove this gradient, while I thought that integrating 40h we should just have less problem.
here I need help! and I would surely have other questions about DBE that I use to solve this kind of problem but who do me things also bisard but everything in his time for now I would like to understand what goes wrong!
if you need special information tell me I will respond quickly.
Thanks again to all! and thanks Pixinsight which is really great software for astronomy!
And sorry for my english ;)
-
If you know of a way to make the XSIF file available, I'll post it
-
with respect to the M81/M82 picture, you know about the Integrated Flux Nebula, right?
the M101 image seems fine, it just needs to be color balanced.
or are you talking about the vertical streaks in the M81/M82 image?
you can upload an XISF to google drive or dropbox or OneDrive and post the link here...
rob
-
yes I know there is IFN lol still happy with 40 Hours!
the problem is the very present chromatic noise, you see, different colors all over the place?
vertical streaks are a problem of banding of a bad dithering ...
ok for the xsif I send it away
-
normally it's good here are the links: https://drive.google.com/file/d/1lBvq0qTiztjsKM5hKlMP5T3C2dmqyTNT/view?usp=sharing
-
yes, now i see what you are talking about, the thumbnails are too small to really see it well.
the CanonBandingReduction script (with highlight protection turned on) can help with the banding (but you have to rotate the image 90 degrees.) however the background is definitely weird, and i see the streaks. i think those are probably caused by hot pixels not being corrected during calibration. you can try turning off Dark Optimization to see if it fixes that, or else run your subs thru CosmeticCorrection to try to eliminate any hot/warm pixels left in the calibrated frames.
it's hard to attack the color un-evenness with DBE because of the IFN...
rob
-
for the banding, I actually used the script and it worked well, but as you say it's the background that has a problem ...
I will try without optimization but I do not think it comes from that.
For cosmetic correction I am of course used on all calibrated images.
I managed to treat it is not the problem, just I would like to understand what is wrong for such an ugly integration ...
The final image is here: https://www.astrobin.com/323598/E/?nc=user
-
I'm not trying to give how to solve the problem but more of a case of showing the problem on a larger scale. Your background noise looks like an alignment problem. I wanted to see the noise without as many distractions from the rest of the image. Here is what I found. On the left side of the image the noise sweeps towards the upper left. In the center it is straight across the image and on the right side it sweeps towards the upper right. If you have an alignment problem I would think there would be no difference if you had 1 hour or 40 hours of integration, the problem would still be there.
Just my thoughts.
Mike
-
just guessing here, but if there is a lot of pattern noise left in the calibrated subs, and they were registered with distortion correction, maybe that's where the noise "directions" come from.
anyway if you already did CC and there's no hot/warm pixels left over in the subs, then perhaps my theory is wrong. but it looks typical to me for DSLR images - for whatever reason there always seems to be some kind of pattern noise left over in DSLR images. the banding is an example of this - in areas of low signal, the image has been corrupted by the camera electronics.
rob
-
Hi rob
yeah the DSLR walking noise is something I deal with from time to time because I don't guide and shoot from a Bortle 7-8 location in high humidity. The LP here is a ..... Generally with mine the noise is in one direction not this semi-circular pattern. When I was playing with the image and looking at the different scales there is electronic pattern noise also but it is way down and didn't seem like an issue but I guess it could be.
Mike
-
looking more carefully at your inverted image, maybe this is just due to field rotation? as you point out the noise seems to make a semicircular sweep. IIRC the effects of field rotation are worse near the pole (could be wrong about that)
rob
-
Now that you tell me, it is true that I have a problem of rotation of field quite recurrent, my setting in station is very good and I guide to the optical divider, however I think to have a big error of cone on the tube. ..
In addition, I have no problem on M33 and M31 and other objects in this region.
Yet an alignment with distortion correction should not solve the problem?
In any case I had never thought of that, it's a point that I will have to look carefully.
How do you advise me to align my brutes?
Thank you all in any case for your time!
-
i think StarAlignment can take care of the rotation with no problems, but what's happening is that the noise is getting distorted by the distortion correction, and then when it is stacked you get these funny moire patterns in the noise.
but is distortion correction really turned on? perhaps try running SA with distortion correction turned off? by 'brutes' you mean "raw frames" right?
rob
-
I am talking about light calibrated sorry.
It's an interesting track so I'll try several tonight and I'll let you know!
Thanks
-
Well, that did not change anything ...
There is a little less light because I did not want to reprocess the 489 lights but the problem is still present after an alignment without correction of distortion.
(https://i.imgur.com/Yj0petRh.jpg)
-
When you have a data set this large you can try creating a custom registration frame.
You have 487 light frames so go through each night and pick 3 subs at the beginning middle and end of the night. You then use the mosaic merge tool in Star Alignment to merge these together. I prefer to use the frame adaptation setting for this. Alternatively you can use register/ union-separate on each sub and use dynamic alignment to combine them.
When you have a mosaic image created from each night you can combine these in star alignment and then either register them into an image or again use mosaic merge in star alignment or dynamic alignment to combine all 3. The final goal here is to have a master reference frame that has not bee cropped to 1 image. It should show all of the different subtle offset you had from re framing and dithering each night. You will register all of your images to this master frame and then integrate them in image integration.
I highly suggest using the linear fit clipping rejection algorithm along with large scale pixel rejection. I probably sound like a broken record because of the past few posts, but use a very small ROI box on an opened image and run image integration. See if you are able to notice a difference in your back ground. After an integration the high clipped pixel map will be filled with hot pixels and the low clipping map will contain the overlapping edges of your frames, you can fine tune this. Large scale pixel rejection will also play a role in how your edges overlap, you cna adjust these to taste as well depending on how much signal loss or gain may have occured in an overlapping area. Its good to crank up the buffer and stack size when you are using an ROI, assuming you have enough ram and a fast processor, be sure to set them back to normal when you are happy with the results and ready to stack everything at once. I suggest you sepnd a long time with ROI in different areas of an image before you finally integrate.
Also, when you have the final integrated image, are you linear fitting the data first?
For dslr especially with this large of an integration you should be splitting the channels, running statistics on each channel at 14bit with normalization off to find the lowest median
now you use the linear fit tool to fit the lowest median channel to the other two channels, and finally recombine the channels using recombination.
Now you can do background neutralization with a small sampled area and determine if further gradient removal is needed. I have much better results with ABE than DBE with thick ifn areas, For me the best solution is to try and have ABE apply very subtle improvements to any gradients so that it does not over correct and darken areas where it suspects a flaw. When you have a good subtle setting dialed in you can run multiple iterations of abe until you have the data where it should be. In your example with excessive noise issues, I seem to notice that abe does this alot, it just seems to be the nature of this tool. I try to avoid using background gradient removal if at all possible.
I shot this region using DSLR about a year ago using the aforementioned technique.
https://www.flickr.com/photos/lmnosunsetdeluxe/33193121913/in/dateposted/