Guidance needed on mosaic problem

[blave549]

< edit: I forgot to stretch the JPEG before attaching it here... Fixed. >

Hopefully what I'm doing wrong or not doing at all will be evident to someone...

I am working on a 2x3 mosaic of the Heart Nebula area in OIII. I've already created a mosaic of the area for Ha, and the results were not only fairly easy to obtain (at least in PI terms  ) but are pretty acceptable to my eye. But, the OIII set is not playing well at all for me for some reason.

Each OIII panel is an integration of six 10 minute subs. I've tried both ABE and DBE on each panel prior to using GradientMergeMosaic, but the mosaic area corresponding to one of the panels -- the IC1795 nebula region -- always ends up having a very different "look" than the other five. I wonder if this is because there's a lot more signal there? I've tried to process each panel's stack identically.

I should note that after many tries I was unable to get the OIII panels to co-register with Star Alignment; I didn't know about some of the more clever techniques (e.g. creating and using an artificial starfield as a reference) until today so I haven't tried any of those things yet, but what I've done instead is use my Ha mosaic as the reference image for alignment, and that seems to work just fine. But I'm not sure if this might be a factor in the problem I'm having.

I am also, courtesy of the PI tutorial videos about mosaics, aware that PixelMath can be used to address edge transition issues, but my problem is so drastic that I'm thinking that there must be a better way to address it.

A JPEG of the OIII mosaic is attached. The Ha equivalent is at http://www.flickr.com/photos/46580576@N07/10046413393/.

thanks,

Dave B.

[Tom OD]

Hi Dave,
Are you using the Ha mosaic as the reference frame in Star Alignment? If so select the Register option at the top of the Star Alignment field. I think selecting Frame Adaption to is needed here. Do this for each O3 frame, so they all get registered to the Ha mosaic you made. Then use Gradient Merge on those Star Aligned (to the Ha frame) O3 frames to make the O3 layer.
I have made large base layer panels and some are similar to what you have shown. One or 2 frames look very washed out compared to the others. I would change the order in which I joined them in Star Alignment. Maybe start with a center frame and then aligning them clockwise about the center. I got very different results trying different order of stitching for the base layer.
Here is some reading for you on this subject, http://pixinsight.com/forum/index.php?topic=5786.msg39509#msg39509

Tom

[georg.viehoever]

1. Artefacts: you need to control for bright stars just on the boarder of 2 images. Remove them in one of them (replace them with black).
2. Problem with panel bottom right: it appears that this panel has a higher contrast than the other ones. GMM only adapts brightness levels at the seams, it does not do any scaling. Did you make sure they have similar contrasts? Apart from that, I have no idea what is going on.
Georg

[blave549]

Thanks for the replies.

Tom: that is exactly the process I used.

But -- I think I've figured out what happened. Apparently when I took the subs that comprise that one high-contrast panel, I somehow selected the Ha filter even though the subs' names have "OIII" in them.

That explains a lot, doesn't it! Doh!

BTW the only reason I figured this out is that I decided to try calibrating the subs (I haven't been using darks or bias frames, as it doesn't seem to make any difference in my results). I fired up the Batch PreProcessing script (normally I register and integrate separately), and noticed that some of the frames were grouped under the "Ha" header!

thanks,

Dave B.

[papaf]

Georg,
suppose one (like myself) has this kind of contrast difference problem, how would one tackle it?
If at all possible, of course....

[georg.viehoever]

Hi Papaf,

Note that blave549 apparently had a problem during capture, which is what should be fixed in the first place.

If you still have the problem of different contrasts:

• In theory, FrameAdaption during registration should do the trick. As I understand http://pixinsight.com/doc/tools/StarAlignment/StarAlignment.html#description_005, this happens between reference image and to-be-aligned image during registration. So if your reference mosaic has been created from your tiles, this obviously will not help.
• LinearFit works similar to frame adaption, so this does not help as well.
• One option would be to make the statistics (mean, variance) compatible manually. A PixelMath expression should do the trick. If this is usually sufficient, I dont know.
• My prefered option in difficult cases would be to do the full processing on each tile separately, with the goal to achieve a "compatible" appearance, including the non-linear HTs. Then apply GMM to to processed tiles. There is a fair chance that this will produce good GMM results - mainly because the contrast in processed images is usually much lower than in linear images , which results in fewer GMM artifacts. It may be necessary to process the resulting mosaic again.

Georg

[papaf]

Thanks Georg,
I had a problem during capture too, in which I started too early because of time constraint and the sky wasn't astronomically ready yet. So the first tile got a significantly brighter background...