Some help please with a widefield mosaic

[AndyKeogh]

I have chosen to create a mosaic of that part of the milky way which was visible from my hotel balcony in La Palma.

I thought it would be a challenge and it certainly has been as it has shown up a lot of limitations with some of the common astro packages

I took some 6x4 and 3x4 mosaics with a canon 5D mk iii & 85mm f1.8 lens mounted on a Gigapan Epic pro which in turn was mounted on an Astrotrac travel system. This gives images around 16x24 degrees fov with ~22mega pixels each

At this stage I am on an extremely steep learning curve with Photoshop and the attached image is before I had learnt to use levels, curves and overcome the 2GB file limit; but it gives a flavour of what I am trying to achive. As you will see there are dark borders introduced by dss. I resolved these by converting all raw files to tiff before dss and that has been sucessful.

My immediate problem is that I am failing to line up each panel with its neighbour. MS Ice, Photoshop photomerge will only align some rows but not the columns.

There is some lens distortion towards the corners,( I shot at f4.0 ) so I have reprocessed some test panels through DxO version 8. I can now align 3 of the 4 panels in a 2x2 matrix. The last panel can be aligned with its neigbour in the same row but the stars do not line up with the row below. At the bottom edge opposite to the join , the error is nearly 400 pixels. this equates to a 4 degree alignment error.

So my guess is that some geometry change will be required beore they are stacked.

My question to the panel is what should I do no next. Giving up is not an option.

TIA

Andy

[ollypenrice]

You are meeting the wonderful world of field curvature. The best software for combining astro images is Registar, which isn't cheap - but you did say that failure was not an option!

Even with Registar you have to be careful, starting with a middle image and co registering to that one. Better still you can use a template - a very wide field image take in a shorter FL lens. Align all your mosaic panels to this rather than to each other.

In all mosaic making projects it is best to crop off the edges which tend to be full of artefacts.

Olly

[AndyKeogh]

Thanks Olly. I have now installed a trial version of Registrar and will go through the example files in the morning. I was using Layers set to difference to find the best alignment then I was going to work out what to remove. As this is my first Mosaic and first session with the 5D I did not know what distortion to expect. Next time I will be more generous with the overlap.

[Psychobilly]

Have you tried the Free Microsoft ICE app?

http://research.micr...groups/ivm/ICE/

I have Registar and it is superb but can be a bit confusing to use (Olly is a Registar Master) ...

I tend to shoot with a generous overlap to allow the edges of the individual panes to be cropped.

Peter...

[foundaplanet]

Is there a general rule of thumb for the percentage of overlap or does that come down to the lens/scope being used.

[Psychobilly]

I tend to go for a minimum of 25%.

[AndyKeogh]

Have you tried the Free Microsoft ICE app?

http://research.micr...groups/ivm/ICE/

Peter...

Thanks Peter, I had tried Microsoft ICE at the begiining of this journey but I seem to recall that it was not very good at the joins. I also thought it only gave 8 bit output.

I have had another look this morning and it does give a 16 bit export, so it is a possibility.

I reprocessed the DSS output files ( which are based on my raw files rather than tiffs so they still had the dark borders ) and I have attached a thumbnail below.

So I may use ICE once I have got a little further down the road with learning to use Photoshop.

Thanks again.

Andy

[foundaplanet]

That is starting to look very good.

[swag72]

I used PTGUI for creating my Heart nebula 3x4 mosaic. You select stars to align on each side and then select the same stars in the image to the side. I found it worked very well. I was only working with a 10% overlap at the time as well.

if you can get it as a trial, it may be worth a go.

[AndyKeogh]

Thanks Olly. I have now installed a trial version of Registrar and will go through the example files in the morning. I was using Layers set to difference to find the best alignment then I was going to work out what to remove. As this is my first Mosaic and first session with the 5D I did not know what distortion to expect. Next time I will be more generous with the overlap.

Initial testing went well and I was able to produce a 2x2 mosaic of my test files.

I then tried a 3x2 mosaic but Registar ran out of memory when I tried to combine the images.

I cannot save any files with the trial. Do I need to combine the files or can I export the aligned file after registration and combine in Photoshop? If so it will still limit the size of Mosaic that Registar can handle.

Are there any plans for a 64 bit version?

Andy

[Psychobilly]

Registar hasn't been updated for many years... It's a pity as although its a bit if a one trick pony it does perform it very well...

It really came into it's own when combining data sets from different camera and optical combinations and working with minimal overlaps...

With ICE it's better if you can apply flats to the individual frames before combining. you might be able to improve them using the lens correction tools in CS...

Peter...

Sent from my GT-P7300 using Tapatalk HD

[AndyKeogh]

I used PTGUI for creating my Heart nebula 3x4 mosaic. You select stars to align on each side and then select the same stars in the image to the side. I found it worked very well. I was only working with a 10% overlap at the time as well.

if you can get it as a trial, it may be worth a go.

Wow what a nice tool Seemed to cope with my large files very well and even exported a psb file to photoshop. Mind you I have a lot to learn before I can understand what is happening in blending layers.

It produced a darker image and seemed to highlight level differences between the rows in the Mosaic. See sample below. The flecks are a watermark to show that I am using a trial.

I like the control the program gives over the stitch points.

I have an issue with how it selects the joins. It seemed to favour hiding the right hand side of each pane with the left hand side of the next pane. In my case the right hand side of each pane had the better star arrangement Is there a setting I missed.

I think I will give it a 23 panel mosaic and see how it copes with 500mega pixels

Thanks Sarah and everyone else it looks like there are a number of options to be explored.

Andy

[swag72]

Glad that you found PTGUI useful, it was recommended on another astro forum for making mosaics. I was quite pleased with it. I didn't really dabble with it much so can't help with any settings about which way it overlays from left to right.

[AndyKeogh]

Well I have carried out a few more tests after googling panoramic stitching software and selecting Kolor Autopano Giga 3.0 as it seems to be a more expensive rival to PTGui.

I have a test 6x4 mosaic with a single 60 second exposure for each panel, Each has been dark subtracted and has had flats applied in DSS but no other processing at this stage and teh total file size is 2.85GB .

Both PTGui and Autopano had difficulty finding match points in about 10 of the images but as they both allowed control points ( their terminolgy for points that match) to be defined I used that to add aboutt 100 in each test.

It probably took about 4 hours in each tool to get all the panels to link up and produce a decent rendering. Thumbnails for each tool are below together with a reference image from Carinasoft's Voyager program. The circles are centered on each of the panels making up the mosaic and there is an equitorial grid in hte background.

The rendering took about 5 minutes once all the control points were in place. Autopano used all the memory that was available ( ~15GB) and kept all cores (8) busy for several minutes

Each tool has a range of projections but I limited this test as I wanted to know if either tool could handle the 2.85GB of input data.

Both produced a photoshop psb file of around 10GB but as they are trials they did not provide the panels in usable layers.

So thanks to everyone for your help but as with most things in life a little more research is required before I decide which tool to acquire. Meanwhile back to learning Photoshop.

Andy

Result from Autopano Giga

Reference view from Carinasoft Voyager

Result from PTGui

[Psychobilly]

it's interesting to see the differences in the geometry of the output from the different programs.

Peter..