Creating multi-panel mosaics, overlap % of the panels and processing?

[ONIKKINEN]

The only target that doesn't fit my setups field of view that im interested in (200mm F4.2 newton, aps-c sensor) is the Andromeda galaxy, which i obviously want to image if the clouds ever go away.

I am using NINA so creating the mosaic itself is nothing but a click away in the framing tool, but i have no idea about the specifics, for example how much should i overlap the frames? My scope being a newtonian will probably have some not-quite-corrected coma and other tracking artifacts at the edges so i assume 10% is not enough. The other unknown is exposure time per panel. In one hand i think i could get away with as low as 30 minutes per panel since i will definitely bin the final picture anyway but is that enough? Ideally i would be able to shoot this in one night to get the conditions as close to eachother as possible per panel, or is this overthinking it? I would like to plan ahead and not spend valuable time outside tinkering with the details.

As for the processing part of actually combining the panels is a complete unknown for me, i know some software can do this but are there any recommendations from people who have done the same? I assume i would roughly process each panel first and then do the combining. The exposure time per panel problem of course goes away if some software can equalize different gradients and background levels from different sessions.

[ollypenrice]

I would say that a 20% overlap will make life easier. Automated mosaic software may well handle a two-panel quite easily but, as any mosaic grows, it becomes increasingly unlikely that automated software will succeed. This kind of software is very competent with daytime images but astro images are massively stretched and present a much bigger challenge.

As for exposure time, how deep do you want to go? I was intrigued by the size of M31 on star charts. It was way bigger on the charts than on most images, so I decided to try a set of 30 minute subs to see if I could find the galaxy's outer reaches as seen on the charts.

For mosaics I use Registar in conjunction with Photoshop.

Olly

 

[ONIKKINEN]

29 minutes ago, ollypenrice said:

I would say that a 20% overlap will make life easier. Automated mosaic software may well handle a two-panel quite easily but, as any mosaic grows, it becomes increasingly unlikely that automated software will succeed. This kind of software is very competent with daytime images but astro images are massively stretched and present a much bigger challenge.

As for exposure time, how deep do you want to go? I was intrigued by the size of M31 on star charts. It was way bigger on the charts than on most images, so I decided to try a set of 30 minute subs to see if I could find the galaxy's outer reaches as seen on the charts.

For mosaics I use Registar in conjunction with Photoshop.

Olly

 

I should have maybe mentioned that i am using an EQM-35PRO, which is not a very nice match (read:nightmare) for the 8inch newton. I will not be going over 60s subs in any scenario and preferably i would go for shorter. I do have a new camera that is still yet to see its first light, but it has 16bit adc, 80%+ QE, 14 stops of dynamic range and practically no noise so im hoping 30s or even shorter exposures bring out the halo at least to some extent. What i was wandering how long the integration time should be per panel with a 2x2, 3x3 or even a 4x4 bin in the final combined picture. I will need at least 3 panels, preferably 4 to get roughly the field of view as in yours.

Yours looks fantastic with the well captured halo much farther out than i usually expect to see, would be happy for a half as good capture as that!

[symmetal]

A point to note is that with an EQ mount there will be image rotation between mosaic frames that differ in RA, so you need to ensure any overlap covers this. The rotation amount depends on the Declination of the target. At Dec 0 there is no rotation and it increases with an increase in Dec being very significant nearer the pole. Here's the effect at around Declination 60

Alan

[ONIKKINEN]

18 minutes ago, symmetal said:

A point to note is that with an EQ mount there will be image rotation between mosaic frames that differ in RA, so you need to ensure any overlap covers this. The rotation amount depends on the Declination of the target. At Dec 0 there is no rotation and it increases with an increase in Dec being very significant nearer the pole. Here's the effect at around Declination 60

Alan

Never heard of the effect before, thanks for the heads up. I think this is the reason my previous shots have all had some sort of field rotation if i shot on different days and didn't quite nail the framing right.

[symmetal]

23 minutes ago, ONIKKINEN said:

Never heard of the effect before, thanks for the heads up.

Yes, It doesn't seem to be mentioned when dealing with mosaics for some reason but becomes clear when you think about it. 😀

If you keep a constant declination and move the scope in RA it follows that line of declination. As shown in the star chart below lines of RA are always at right-angles to lines of Declination, so to maintain that condition the camera has to rotate its orienation too, being locked to RA. You could always rotate the camera orientation for each mosaic image but it's too much trouble to work out how much. 🤔

Alan

[ollypenrice]

 

54 minutes ago, symmetal said:

A point to note is that with an EQ mount there will be image rotation between mosaic frames that differ in RA, so you need to ensure any overlap covers this. The rotation amount depends on the Declination of the target. At Dec 0 there is no rotation and it increases with an increase in Dec being very significant nearer the pole. Here's the effect at around Declination 60

Alan

I'm not sure that I'd call this field rotation. I think the problem is geometric. The sky we are photographing is, in effect, seen on the inside of a sphere but our final image will appear on a flat surface. This is a familiar problem in cartography. There are different projections of the 3D earth onto 2D paper maps, each with its own strengths and weaknesses. When creating a large mosaic you need to decide on your field geometry. If you just pick a random panel from your set the software will treat that panel's geometry as definitive and all other panels will work outwards from that one. The best you can do in this case is start with a panel in the centre of your image.

A better idea is to take a widefield image covering all of your intended mosaic, centered on the same point. This can be resampled upwards to the size of the intended mosaic (it will look terrible but that doesn't matter) and it will become your registration template for all your mosaic panels. Your final mosaic will have the field geometry of your widefield image.

Olly

[tomato]

It’s very tempting to try and fit extended objects like M31 in the minimum of panels but the downside is you sometimes have to orientate the object to some odd angles which can detract from the finished image.

Here are two examples, using the same telescope and camera, a 6 panel mosaic which gave a ‘widescreen’ M31, and a 12 panel version which put the galaxy in the more traditional diagonal orientation.

The big advantage was the 6 panel version was captured as 6x1 hr panels in a single clear night in October, the 12 panel version took close on 3 months to complete.

Astro Pixel Processor will do a decent job of combining the panels, especially if they are taken under similar conditions, this was used on the 6 panel version. However, removing joins and gradients taken over many sessions can be a real challenge, by far the best results I have obtained have been with the Photometric Mosaic script available in Pixinsight. 
 

Best of luck with your M31 project, I will be attempting another mosaic on this iconic target shortly (or when the weather actually permits).

[ONIKKINEN]

2 minutes ago, tomato said:

It’s very tempting to try and fit extended objects like M31 in the minimum of panels but the downside is you sometimes have to orientate the object to some odd angles which can detract from the finished image.

Here are two examples, using the same telescope and camera, a 6 panel mosaic which gave a ‘widescreen’ M31, and a 12 panel version which put the galaxy in the more traditional diagonal orientation.

The big advantage was the 6 panel version was captured as 6x1 hr panels in a single clear night in October, the 12 panel version took close on 3 months to complete.

Astro Pixel Processor will do a decent job of combining the panels, especially if they are taken under similar conditions, this was used on the 6 panel version. However, removing joins and gradients taken over many sessions can be a real challenge, by far the best results I have obtained have been with the Photometric Mosaic script available in Pixinsight. 
 

Best of luck with your M31 project, I will be attempting another mosaic on this iconic target shortly (or when the weather actually permits).

I agree that i prefer Andromeda to be a bit diagonal. Its a weird opinion since what points to what direction is completely arbitrary and all opinions and orientations are "correct". This is however not a problem since i will just rotate the entire imaging train to 90 degrees to reach this orientation.

Something like this is my plan. It could use a few more panels but if i get this to work and all frames to be more or less on point it would be good for me. Local weather is dreadful and i don't have a backyard, so 10+ hour projects are not something im looking for right now.

Astro pixel processor does look pretty nice. Even comes with a free trial and a yearly rent period. I am aware of pixinsight too but that software looks like it was written by aliens, for aliens, with alien language. Complete gibberish to me when I've tried looking at some tutorials.

[symmetal]

36 minutes ago, ollypenrice said:

I'm not sure that I'd call this field rotation. I think the problem is geometric. The sky we are photographing is, in effect, seen on the inside of a sphere but our final image will appear on a flat surface

No it's not field rotation as one image doesn't rotate during capture.

The cartography problem of fitting a sphere to a flat surface is the same no matter which point on the sphere is chosen to be the centre of your flat surface.

With an EQ mount , at Dec zero there is no mosaic panel image rotation if RA is changed. The effect is a consequence of the mount design rather than a surface mapping problem. 😉

Alan

Edited August 28, 2021 by symmetal

[ONIKKINEN]

19 minutes ago, ollypenrice said:

 

I'm not sure that I'd call this field rotation. I think the problem is geometric. The sky we are photographing is, in effect, seen on the inside of a sphere but our final image will appear on a flat surface. This is a familiar problem in cartography. There are different projections of the 3D earth onto 2D paper maps, each with its own strengths and weaknesses. When creating a large mosaic you need to decide on your field geometry. If you just pick a random panel from your set the software will treat that panel's geometry as definitive and all other panels will work outwards from that one. The best you can do in this case is start with a panel in the centre of your image.

A better idea is to take a widefield image covering all of your intended mosaic, centered on the same point. This can be resampled upwards to the size of the intended mosaic (it will look terrible but that doesn't matter) and it will become your registration template for all your mosaic panels. Your final mosaic will have the field geometry of your widefield image.

Olly

I do have a 55-250mm kit lens and a Canon 550D, would this work for the widefield template or is it a better idea to take a centered frame with the main imaging scope?

[ollypenrice]

3 minutes ago, ONIKKINEN said:

I do have a 55-250mm kit lens and a Canon 550D, would this work for the widefield template or is it a better idea to take a centered frame with the main imaging scope?

I would take a widefield with your lens, upsize it to the scale of the projected mosaic and use it as a template. I use Registar for this kind of task.

Olly

[ollypenrice]

11 minutes ago, symmetal said:

No it's not field rotation as one image doesn't rotate during capture.

The cartography problem of fitting a sphere to a flat surface is the same no matter which point on the sphere is chosen to be the centre of your flat surface.

With an EQ mount , at Dec zero there is no image rotation if RA is changed. The effect is a consequence of the mount design rather than a surface mapping problem. 😉

Alan

You may be right but I don't follow this argument. It seems to me that, for a given moment, all mounts are identical. They point somewhere, centered upon co-ordinates x and y.  The camera centered upon these co-ordinates will render a curved field onto a flat chip in accordance with the optics in use. You could take this image quickly (in principle) with a fixed tripod or an EQ mount or an alt-az but it will remain precisely the same image with precisely the same field curvature. Give this image to a standard mosaic software and it will extend that image's curvature outwards into the rest of the mosaic.  In my view the mount design has zero importance because, if/when we have all the photons we need in one second, we won't need tracking mounts but we'll have our picture, and the geometry of that picture will be exactly as it was in the tracked version. If this were not true the tracked version would not have round stars. The whole point of an EQ mount is that it gives the same result (only deeper) than a short snapshotun-tracked.

Olly

[symmetal]

3 hours ago, ollypenrice said:

You may be right but I don't follow this argument. It seems to me that, for a given moment, all mounts are identical. They point somewhere, centered upon co-ordinates x and y.  The camera centered upon these co-ordinates will render a curved field onto a flat chip in accordance with the optics in use. You could take this image quickly (in principle) with a fixed tripod or an EQ mount or an alt-az but it will remain precisely the same image with precisely the same field curvature. Give this image to a standard mosaic software and it will extend that image's curvature outwards into the rest of the mosaic.  In my view the mount design has zero importance because, if/when we have all the photons we need in one second, we won't need tracking mounts but we'll have our picture, and the geometry of that picture will be exactly as it was in the tracked version. If this were not true the tracked version would not have round stars. The whole point of an EQ mount is that it gives the same result (only deeper) than a short snapshotun-tracked.

Olly

You're right Olly that a camera on any mount can create a mosaic by pointing the camera at the celestial coords corresponding to the centre of each panel of the mosaic.

My argument is only concerning the resultant orientation of the camera for each panel. Here's a Photoshop drawing showing the extreme option of a 4 panel mosaic centred on the Celestial Pole using 3 different mounts which illustrates the point. 😉 The camera isn't touched between panels.

For high declination mosaics EQ mounts present extra challenges, although of course don't suffer from field rotation like the other two will. 🙂

Imaging software which generate mosaic panel coords, (at least the ones I've looked at) don't consider this feature of EQ mounts and assume mosaics are constructed as if it's just a camera on a tripod. 

Alan

 

Edited August 29, 2021 by symmetal

[ollypenrice]

On 28/08/2021 at 23:25, ONIKKINEN said:

I do have a 55-250mm kit lens and a Canon 550D, would this work for the widefield template or is it a better idea to take a centered frame with the main imaging scope?

No, you can just use the camera lens centered on the middle of the intended mosaic. Whatever the field curvature of that lens is, it will then define the curvature of the mosaic. You'll be using it as a template. At least, that's how I would do it using Registar for the mosaic construction. I'd take the lens shot then work out the size of the final mosaic and re-size the lens shot to that size. (You might consider making the final mosaic a little smaller since getting quality good enough for full size presentation can be difficult and time consuming at capture.)  In Registar you can then register each panel to the template, discard the template and combine all the registered panels.

There are other ways. I've post-processed a mega-mosaic captured by a client and using a template he made in Astro Pixel Processor from individual panels. APP did a good job of the template and field geometry but couldn't match up the panels seamlessly so I kept the template and then worked as above using Registar.

Olly

[ONIKKINEN]

On 29/08/2021 at 00:07, tomato said:

It’s very tempting to try and fit extended objects like M31 in the minimum of panels but the downside is you sometimes have to orientate the object to some odd angles which can detract from the finished image.

Here are two examples, using the same telescope and camera, a 6 panel mosaic which gave a ‘widescreen’ M31, and a 12 panel version which put the galaxy in the more traditional diagonal orientation.

The big advantage was the 6 panel version was captured as 6x1 hr panels in a single clear night in October, the 12 panel version took close on 3 months to complete.

Astro Pixel Processor will do a decent job of combining the panels, especially if they are taken under similar conditions, this was used on the 6 panel version. However, removing joins and gradients taken over many sessions can be a real challenge, by far the best results I have obtained have been with the Photometric Mosaic script available in Pixinsight. 
 

Best of luck with your M31 project, I will be attempting another mosaic on this iconic target shortly (or when the weather actually permits).

I have just tried combining the stacks from DSS to do the mosaic in APP. There are obvious lines between the stacks. Should i do all of the process in APP? I tried but it had already taken an hour of registration and it was nowhere near done so i just canceled it, is this normal for APP? For comparison i dont think the process took more than 20 minutes from unloading my memory card to having all 4 panels stacked in DSS.

[tomato]

I would do everything in APP. Calibrate and stack each panel first, then combine the stacked panels to create the mosaic. You need to experiment with the Multi Band Blending and Local Normalisation Correction settings to achieve the best seamless result.

Take a look at this excellent guide by Sara Wager:

 https://www.astropixelprocessor.com/how-to-create-a-mosaic-in-easy-steps-by-sara-wager/

I don’t know the spec of your processing PC, but the processes can take a while  on a lower spec machine.

[ONIKKINEN]

On 01/09/2021 at 09:28, tomato said:

I would do everything in APP. Calibrate and stack each panel first, then combine the stacked panels to create the mosaic. You need to experiment with the Multi Band Blending and Local Normalisation Correction settings to achieve the best seamless result.

Take a look at this excellent guide by Sara Wager:

 https://www.astropixelprocessor.com/how-to-create-a-mosaic-in-easy-steps-by-sara-wager/

I don’t know the spec of your processing PC, but the processes can take a while  on a lower spec machine.

Thanks for the link, mostly helpful. But the example in the tutorial has 29 subs and doesn't do star analysis or registration, which are the phases that took my 250 subs around 4 hours. I just found it weird that APP hangs on to the process for so long, i had to leave the PC running as i went to sleep because it had already taken at least 6 hours. Wondering if i did something wrong? Mostly just followed the recommendations in APP. My PC is not afraid of a slight breeze, its an overclocked 6700K with decent DDR4 RAM, if that means anything to you, so weird that it took so long.

Anyway as a proof of concept i combined 5 panels worth of data into one and got it working fairly well. There are seams but they are far less obvious than i expected. Also the middle parts are missing the blue halo of young hot stars as it has the least data. I think i got the hang of it now, just need to set aside a full processing day to get this done whenever i return to a mosaic project.

[symmetal]

22 hours ago, ONIKKINEN said:

Thanks for the link, mostly helpful. But the example in the tutorial has 29 subs and doesn't do star analysis or registration, which are the phases that took my 250 subs around 4 hours. I just found it weird that APP hangs on to the process for so long, i had to leave the PC running as i went to sleep because it had already taken at least 6 hours. Wondering if i did something wrong?

Stacking in APP does take ages with lots of large files. I like to assume that at least it's doing a good job. 😀 I only use it for stacking at the moment and now run APP and Startools on my gaming PC with an i9 processor, 32GB ram and RTX2070 graphics card. Stacking around 300 images from my ASI6200 with 116MB image files can still take around 2 hours. On my usual i7 PC it would likely take around 5 to 6 hours.

I used to use Astroart for stacking, which is much quicker, but found I would sometimes get strange background banding artifacts when using Develop in Startools. With APP stacking I don't, so use that now and put up with the wait. 🙂

Very impressive looking mosaic by the way. 

Alan