Large Mosaic Best P...
 
Share:
Notifications
Clear all

31 July 2020 - Comet Registration video tutorial using APP 1.083-beta1 released.

30 July 2020 - APP 1.083-beta1 has been released introducing Comet processing! This 1st beta has comet registration. The stable release will also include special comet integration modes.

9 July 2020 - New and updated video tutorial using APP 1.081: Complete LRGB Tutorial of NGC292, The Small Magellanic Cloud by Christian Sasse (iTelescope.net) and Mabula Haverkamp

2019 September: Astro Pixel Processor and iTelescope.net celebrate a new Partnership!

Large Mosaic Best Practise?  

  RSS

(@james474789_app)
Hydrogen Atom Customer
Joined: 1 month ago
Posts: 2
September 27, 2020 23:06  

I am working on a 48 panel mosaic and not entirely sure how best to complete preprocessing. For proof of concept I have a single sub of each panel and have loaded this into APP. I changed the stars target to 1300, registration mode to mosaic and followed the guidance other than disabling 'same camera and optics'.

This is taking a long time to integrate the mosaic (i7-6700k with 32GB) and if I enable LNC + MBB it takes most of a day to build the mosaic (all 900MP of it!). My problem comes when I try to do this for real. My understanding of best practise is to integrate each panel individually and then integrate the mosaic. This means running through the small integrations 48 times which seems like torture!

1. Is this still the best approach or is there a better alternative?

To complicate this further I already have additional deeper integrations I want to integrate into this mosaic; some of these are also mosaics (up to 6 panels). My testing so far has only worked when I integrate the deeper mosaics into the relevant panels of the 48 panel mosaic, otherwise it fails to register?

2. At what point should I attempt to integrate these?

 

Any other advice?

 

Thanks,


ReplyQuote
Topic Tags
(@vincent-mod)
Quasar Admin
Joined: 3 years ago
Posts: 2469
September 28, 2020 12:08  

It's best to create panels first indeed, but these panels can be bigger than you have them as long as the overlap is at least close to 10% I'd say. Such a big mosaic still does take quite some time indeed, that's the nature of it I think. I think I need some screenshots of how the panels look and what you try to add to the already existing data.


ReplyQuote
(@james474789_app)
Hydrogen Atom Customer
Joined: 1 month ago
Posts: 2
October 3, 2020 14:03  
2020 10 03

Screenshot of the mosaic attached. This is proving to be a nightmare and my i7 6700k with 32GB ram isnt able to process the source images. I also struggled with registration in APP so I ended up doing part of the processing in PI and part in APP.

1. I batch integrated the HA panels in PI.

2. I batch integer resampled this (bin x3) in PI to reduce the file sizes.

3. I registered the panels in PI using Mosaic by Coordinates. I could not get the 48 bin3 panels to register in APP.

4. I integrated the panels in APP with LNC and BPP. This produced a halfway decent result, but there are blank areas.

 

Link to the 48 bin3 source files:

https://1drv.ms/u/s!AtooZYq7lKxzgWsP_egr7ZBIgnFv?e=3DNOP8

 

Any advice appreciated on how best to process this. In the meantime I am waiting on the release of the successor to the 3950x because this entire process is killing my i7.


ReplyQuote
(@vincent-mod)
Quasar Admin
Joined: 3 years ago
Posts: 2469
October 5, 2020 18:10  

You can start with making panels that cover like 4 of the individual panels in the mosaic. Then make a mosaic of those, should be easier on your system. When that doesn't work I'll download the data to have a look as well (currently not having access to the best wifi so have to be careful).


ReplyQuote
(@tkottary)
Molecular Cloud Customer
Joined: 2 years ago
Posts: 5
October 13, 2020 03:23  

@james474789_app I recently did 21-panel mosaic of 2600mc pro files(51 overall with overlaps). All of these 21 panels were integrated, DBE, cropped to remove stacking artifacts, and saved separately.

With APP, enabled mosaic mode, calibrated projective, and equirectangular or Mercator projection. LNC and MBB you will have to play with it to get suitable results. I end up with an 8th degree and  6 iterations LNC  to get a usable result. Overall it took 1.5 days on my i6700k process with 64gb. Memory staying around 40gb utilized .


ReplyQuote
Share: