Mosaic help please!

@ elgol,

I have been waiting for some updates and found some time to do this for myself which Mabula succeeded in months ago. Some things changed and I cannot get to the point of stitching the panels already failing from the beginning. What is really frustrating is that registration works for hours to then telling me something is wrong.

So, after half a year of waiting, I ask for help!

I checked my reocrds, a workflow to register your mosaic was sent to you 6 moths ago:

Hi Stefan,

I hope this will get you started:

I dividided all your lights into 12 different panels, maybe there are less, but that's not so important.
So 12 panels with a different field of view than the other panels.

I processed each of the 12 panels to a single integration file.

I calibrated all frames with the masterbias, masterflat and BPM.

I am not quite sure that the provided masterflat was for the 135mm or 400 mm focal lenghts.

Before starting with 3) analyse stars, I double checked the calibration with the l-calibrated image viewer mode, for some of the mosaic panels, i didn't use the masterflat. In those cases, the calibration without the masterflat was better, probably this had to do with the different focal lengths.

I never used the masterdark.

The integrations of all 12 panels:

integrated using LNC 1st degree 3 iterations, MBB 5% and median integration without outlier rejection

For some panels, after normalization, I removed a couple of lights due to guiding errors, shown as frames with low star counts and low quality relative to the other frames, so easy to pick them out.

Then the 12 panel mosaic:

load the 12 integrations as lights:

3) star analysis
detect above noise level kappa 025
maximum stars 7000

4) quadrilaterals
scale start 1
scale stop 10
use dynamic distortion correction
disable same camera and optics
distortion margin 0.01
overlapping frames 9
mosaic mode
projective model

mosaic was registered with precision of 0.97 pixel for 13062 stars
This high RMS (usually less then 0.5 pixel) is most likely due to the range in focal lengths between the 135 and 400mm objectives, the reference was chosen automatically as one of the 400mm frames.

5) normalize
advanced mode

6)
equal weights
integrate average
mode full
lnc 4th degree 5 iterations
MBB 15%
no outlier rejection
lanczos 3 with no under/overshoot enabled

test mosaic first made with scale 0.3, which means that mosaic is (0.3/1.0)x the size off the full size and the image scale is 1.0 /0.3 = 0.333x the image scale of the reference.

This test looks awesome

Create full mosaic with scale 1.0 and 10 iterations 4th degree LNC...

I am running the mosaic again now, with APP 1.061, since several things changed in the last 6 months, to verify what settings you need in APP 1.061.

Hang on,

Mabula

@elgol,

I have 10 panels from your data. I created these 6 months ago form your individual subs. These are off course different than your 10 panels, so I will look at those as well.

I tried the mosaic in APP 1.061, and in less than 15 minutes, I re-created your mosaic with more or less the same settings as indicated in the workflow. Perhaps your initial panels have problems, so I need to have a look to see if I can register those as well.

This was just a test run, so I integrated without LNC and MBB and scale factor 0.5x

ALL the settings that I used to Register the mosaic in APP 1.061:

3) Analyse stars

all default, except automatic minimum #stars target, I set this to 2500. (it automatically sets maximum star target to 5000)

In the 10 mosaic panels, the star counts were about 2000-5000.

4) Register

scale start 5

scale stop 10

enable dynamic distortion correction

disable same camera and optics

kept distortion margin on 0.05

registration mode mosaic

registration model projective.

It took only 10 minutes give or take on my machine with 32 cpu threads. So on a machine with only 4 threads, it might take 1 hour give or take.

I will now look at your panels to check/verify if these are okay to start with.

Mabula

Hi @elgol and everyone else,

This superb data by Stefan, will feature in a new Video Tutorial for mosaicing using multi-focal length data 😉 I will work on this the coming days.

Mabula

Thank you Mabula, 

that will be a great help in the increasing complexity of your powerful &  growing tool 

Stefan