Mosaic sure can do interesting things

[Nocturnal]

I was trying to stich two images together when mosaic came up with these two variants. The only difference was that in one case I had the brightness relations unchecked.



StarAlignment: Processing view: NGC4438_West
NGC4438:
Extracting CIE L* component: 100%
Structure map: 100%
Detecting stars: 100%
1667 stars found.
NGC4438_West:
Extracting CIE L* component: 100%
Structure map: 100%
Detecting stars: 100%
1788 stars found.
Matching stars...
569 putative star pair matches.
Performing RANSAC...
9 star pair matches in 225 RANSAC iterations.
Summary of model properties:
Inliers     : 0.016
Overlapping : 0.197
Regularity  : 0.434
Quality     : 0.324
Root mean square error:
?RMS  :  0.705 px
Average RMS error deviation:
?RMS  :  0.372 px
Peak errors:
?xmax :  1.278 px
?ymax :  0.898 px
Transformation matrix:
     -0.5072     -0.0359  +1407.6933
     -0.3146     -0.5332  +1012.4631
     -0.0004     -0.0002     +1.0000
Generating mosaic image: NGC4438_West_mosaic
Homographic Projection / Bicubic Spline Interpolation, c=0.30: 100%
Registration successful.
10.372 s

StarAlignment: Processing view: NGC4438_West
NGC4438:
Extracting CIE L* component: 100%
Structure map: 100%
Detecting stars: 100%
1667 stars found.
NGC4438_West:
Extracting CIE L* component: 100%
Structure map: 100%
Detecting stars: 100%
1788 stars found.
Matching stars...
606 putative star pair matches.
Performing RANSAC...
8 star pair matches in 205 RANSAC iterations.
Summary of model properties:
Inliers     : 0.013
Overlapping : 0.293
Regularity  : 0.493
Quality     : 0.342
Root mean square error:
?RMS  :  0.863 px
Average RMS error deviation:
?RMS  :  0.437 px
Peak errors:
?xmax :  1.573 px
?ymax :  0.892 px
Transformation matrix:
     -0.3741     -0.0018  +1021.7580
     -0.5746     -0.3385  +1712.3293
     -0.0004     -0.0000     +1.0000
Generating mosaic image: NGC4438_West_mosaic1
Homographic Projection / Bicubic Spline Interpolation, c=0.30: 100%
Registration successful.
10.683 s


Switching to 2D projection made PI crash. Woops!

[Nocturnal]

16b versions of the files uploaded here: http://cid-a93625fef5ca95fb.skydrive.live.com/browse.aspx/PixInsight/Mosaic%5E_01?uc=2

[Niall Saunders]

Sander - no included image in your first post . . . . ?

[Nocturnal]

What do you mean Niall? I see a screenshot, you don't?

[Niall Saunders]

Hi Sander,

OK - I have just checked here using a different PC, and I do now see the screenshot.

Which just confirms all the problems that I am having are due to the PC at home - so time to dump the web-browser, and re-install.Oh the joys of Internet Explorer 

I can't actually help you with the mosaic problem - I haven't been able to acquire two images that 'could' be stitched together (clouds, then full moon, then snow, the freezing fog, then more snow, then clouds - and guess what, it will soon be full moon again  )

Cheers,

[Nocturnal]

Thanks for looking Niall, sorry about your bad weather

[Juan Conejero]

Hi Sander,

Yes, this is a known issue that is now completely fixed.

The problem is that I still haven't published the new version of StarAlignment. You'll love it: you no longer have to define previews as SA is now able to find intersections between mosaic frames automatically (through a FFT-based registration algorithm). Previews are only necessary when there are significant differences in scale (which is infrequent with mosaics), or when the overlapped area is extremely small.

Just a bit of patience. I'll try to upload updated ImageRegistration modules tomorrow.

In the meanwhile, I am downloading your images right now. Surely they will be doable with SA by tuning parameters. I'll let you know.

[Juan Conejero]

Hi again Sander,

The new StarAlignment with default parameters (mosaic created in 14 seconds):

http://forum-images.pixinsight.com/legacy/mosaic-sander/mosaic-ngc4438-sander.jpg

Here is the console information:

Code: [Select]

StarAlignment: Processing view: NGC4438_West_16b
NGC4438_16b:
Extracting CIE L* component: 100%
Structure map: 100%
Detecting stars: 100%
1665 stars found.
NGC4438_West_16b:
Extracting CIE L* component: 100%
Structure map: 100%
Detecting stars: 100%
1787 stars found.
Looking for intersection...
Intersection:
R1  : { 2695,   57, 2995, 1974}   160 stars
R2  : {    4,   17,  304, 1934}   197 stars
?21 : 0°
Matching stars...
147 putative star pair matches.
Performing RANSAC...
128 star pair matches in 69 RANSAC iterations.
Summary of model properties:
Inliers     : 0.871
Overlapping : 1.000
Regularity  : 0.925
Quality     : 0.838
Root mean square error:
?RMS  :  0.889 px
Average RMS error deviation:
?RMS  :  0.359 px
Peak errors:
?xmax :  1.952 px
?ymax :  0.259 px
Transformation matrix:
     +1.0251     -0.0070  -2749.3573
     +0.0142     +1.0231    -80.1930
     +0.0000     +0.0000     +1.0000
Generating mosaic image: NGC4438_West_16b_mosaic
Building 2D surface interpolation grid: 100%
2-D Surface Spline / Bicubic Spline Interpolation, c=0.30: 100%
Registration successful.
13.901 s
As you see, SA has found the intersection between both images without problems. The resulting registration model is quite good (0.8 quality). It could be further improved by including more stars (tuning star detection parameters).

[Niall Saunders]

Hi Juan,

Whilst we are on this subject, could you possibly shed some light as to what the elements of the 3x3 Transformation Matrix represent?

Cheers,

[Nocturnal]

Thanks Juan! Will start working on DBE'ing the images so they're ready to go.

[Juan Conejero]

Keep them flattened, as I'll try to release the new SA very soon. I've got a severe cold so I'll be somewhat slow, though...

[Nocturnal]

By flattened you mean linear, right?

[Juan Conejero]

Sorry for the incorrect word (flattened is obviously a verbal form, not an adjective).

Flat (as for illumination profile), and naturally linear, I mean

cof, cof, ... 

[Nocturnal]

Hi Juan,

I uploaded the cloned preview I used:

http://www.tungstentech.com/files/SH2_240_preview05.tif

Snorting salty water through your nostrils helps make you feel better when you have a cold. In fact it can reduce the chances of getting a cold as it flushes cold/flu viri out of your nose before they settle. I use a Neil Med bottle, not sure if it's available in Spain. My dad used to let us snort salty water from a small spoon, that worked too but I prefer the bottle.  http://www.neilmed.com/usa/sinusrinse.php  Get better soon!

[Nocturnal]

Hi Juan,

any word on getting this new SA module out to us? I'd like to give this image anyother try.