My first attempt at building a mosaic so it would be good have a critical appraisal please re: any improvements or suggestions. I chose the SMC as it is well placed at the moment to capture a good amount of data each night. There are some good objects coming into play now which I would like to capture some four panel mosaics on, hence the practice runs on the SMC.
This is a two panel image with HaRGB. Approximately two hours of data per filter per panel. The area is the north east section of the SMC. The panels were assembled in Pixinsight and used tools such as the "Mosaic by coordinates" script and the GradientMergeMosaic process.
In my first attempt I tried assembling the R, G and B filters into separate two panel mosaics and then combing them into a single panel RGB. This didn't work as I finished up with a nasty colour gradient. I then tried assembling the RGB in each panel first and completing DBE, PCC on each before combining and this seemed to work ok.
It was in interesting learning experience going through the process I have to say. The image files finish up being VERY BIG and the PI processes take a while to run through the hoops. And this is only a two panel mosaic. It must be quite a struggle throwing around the panels for larger mosaics.
That turned out pretty well. A bit dark in the background though.
Mosaics are hard work. What worked for me was to take each panel the same time each night (don't try to do more than one panel in a night as different positions in the sky will give different brightnesses). Same everything.
Then blend them in a panorama program. These days Photoshop CC does a reasonable job. Its worth a go. I use PT Gui and its quite reliable.
Microsoft ICE is popular and free.
The hardest thing can be subtle colour differences that show up when blended. Photoshop's desaturate/saturate brush, burn/dodge can be helpful to adjust the borders. Also the select colour tool set to neutrals can be helpful in adjusting subtle colour shade differences.
One word of advice with mosaicing the SMC -- use /huge/ overlaps between each panel. The closer you get to the pole, the greater the overlap needs to be. I learnt the hard way a few years ago where I threw out 60 hours of data because there wasn't enough overlap between the frames. Nowadays, I build a scaffold first and proceed only when I'm satisfied there's enough overlap.
That turned out pretty well. A bit dark in the background though.
Mosaics are hard work. What worked for me was to take each panel the same time each night (don't try to do more than one panel in a night as different positions in the sky will give different brightnesses). Same everything.
Then blend them in a panorama program. These days Photoshop CC does a reasonable job. Its worth a go. I use PT Gui and its quite reliable.
Microsoft ICE is popular and free.
The hardest thing can be subtle colour differences that show up when blended. Photoshop's desaturate/saturate brush, burn/dodge can be helpful to adjust the borders. Also the select colour tool set to neutrals can be helpful in adjusting subtle colour shade differences.
Even so mosaics are for the brave!
Greg.
Thanks Greg, I was wondering if it was a smidge dark. Agreed that mosaics can be hard work (but worth the effort of course).
Thanks for the recommendation on lots of overlap. With this one I used a 20% overlap which seemed to work fine. This still allowed me to trim off some edges to tidy thing up.
Am I going to do the rest Well, we have pondered this at length I must admit. I estimate that I would probably need maybe 8 panels to complete the job properly. Dare I enter this journey? Am I prepared to make this commitment? We shall continue to ponder this matter.
That's just gorgeous. The HII regions look brilliant. Well done!
Really hoping that you can complete this project and that you'll show us the result.
We agree that as far as possible one wants to do each panel of a mosaic under the same conditions. Close to new moon at a really dark site with no high altitude cloud is really helpful. Close to the meridian is really helpful. Really bad is bright moon in one frame but not another, light pollution or moonlight bouncing off high altitude cloud in one panel but not another is bad. Low down in the murk in one frame and at the zenith in another is really bad.
We calibrate to set the black point of the histogram (the leftmost point of the histogram where the frequency is nonzero) to be identical in each frame. As others have mentioned, a big fat overlap really helps.
The absolute killer is if you have different, incompatible gradients in different frames, e.g. moon on the east in one panel, moon on the west in another, or horizon glow to the south of one panel but not in another. If you do have different gradients in each panel, no amount of overlap, no practicable amount of feathering, will get rid of the visible joins. It is far better to take the panels under similar conditions. If that isn't possible, then in an object like this, where you don't have continuous nebulosity throughout the object, you can reduce gradients on each panel before trying to combine the panels.
One other tip: if you're running out of memory, it might be necessary to do the mosaic in say three sub-mosaics. Fully and seamlessly assemble each sub-mosaic. Then combine the three sub-mosaics. This technique will only work if there are no incompatible gradients.
That's just gorgeous. The HII regions look brilliant. Well done!
Really hoping that you can complete this project and that you'll show us the result.
We agree that as far as possible one wants to do each panel of a mosaic under the same conditions. Close to new moon at a really dark site with no high altitude cloud is really helpful. Close to the meridian is really helpful. Really bad is bright moon in one frame but not another, light pollution or moonlight bouncing off high altitude cloud in one panel but not another is bad. Low down in the murk in one frame and at the zenith in another is really bad.
We calibrate to set the black point of the histogram (the leftmost point of the histogram where the frequency is nonzero) to be identical in each frame. As others have mentioned, a big fat overlap really helps.
The absolute killer is if you have different, incompatible gradients in different frames, e.g. moon on the east in one panel, moon on the west in another, or horizon glow to the south of one panel but not in another. If you do have different gradients in each panel, no amount of overlap, no practicable amount of feathering, will get rid of the visible joins. It is far better to take the panels under similar conditions. If that isn't possible, then in an object like this, where you don't have continuous nebulosity throughout the object, you can reduce gradients on each panel before trying to combine the panels.
One other tip: if you're running out of memory, it might be necessary to do the mosaic in say three sub-mosaics. Fully and seamlessly assemble each sub-mosaic. Then combine the three sub-mosaics. This technique will only work if there are no incompatible gradients.
So far, your image looks magnificent!
Best,
MnT
Thanks for the encouragement MnT. I particularly like the reference to setting consistent black points in each frame. A good suggestion. At least with the SMC it is conveniently placed to track the object for a few hours each session so there should be good opportunities to collect data.
and the PI processes take a while to run through the hoops. And this is only a two panel mosaic. It must be quite a struggle throwing around the panels for larger mosaics.
Well, we have pondered this at length I must admit. I estimate that I would probably need maybe 8 panels to complete the job properly. Dare I enter this journey? Am I prepared to make this commitment? We shall continue to ponder this matter.
