Full sky mosaic. Test of MosaicByCoordinates

[Andres.Pozo]

A few days ago I took a set of widefield images for testing the script MosaicByCoordinates. Since I wanted only to do a quick test I didn't go to a nice place so I shot these images between buildings in the middle of a village.

I used a Canon 60D with a Canon 15-85mm lens at 15mm and F3.5. The camera was on a tripod so I had to use a short exposure (10") and a high ISO (6400). The mosaic is composed by four tiles with 53 images in total. The result covers about the 60% of the visible sky.

The process has been this:

• Calibrate the images with dark frames using BatchPreprocess. I tried to use flats, but they weren't good enough.
• Register the images using StarAligment with the distortion correction activated.
• Integrate the images of each tile.
• For each tile: DBE, SCNR(green), BackgroundNeutralization, ColorCalibration
• Solve the coordinates for each tile using ImageSolver. I used an experimental version of ImageSolver which is much more robust that the published one when solving images with big geometric distortions. I will publish it when I talk a couple of things with Juan. With the new version I achieved very good coordinates for the four tiles. This is essential for the next step.
• Using MosaicByCoordinates I registered the four tiles with an Orthographic projection. It shows the mosaic as if it would had been taken with a fish eye lens. This projection is not yet available in the published version.
• Using GradientMergeMosaic I merged the four tiles. I stretched the tiles before GMM since this way it seems to produce better results.
• Since GradientMergeMosaic loses the keywords of the tiles, I copied the keywords of one of the registered tiles to the merged mosaic using FITSHeader. Since the geometry is the same, the coordinates are still valid.
• Final curves adjustments.

This method produces a geometric correct mosaic in any chosen projection and orientation. The result can be annotated with very high precision.

The objective was to prove that MosaicByCoordinates can create a full sky mosaic. I didn't try to get a "beautiful" image. This is the result:

Click on the image for a high resolution version in AstroBin (5949x6636)

This is an annotated version:

Click on the image for a high resolution version in AstroBin (5949x6636)

These are a couple of crops at scale 1:1

[msmythers]

Andres

Very, very impressive image. Your new experimental version of ImageSolver looks like it's working great. Looks like something I could have used for my 15 panel that I finished last month.
http://pixinsight.com/forum/index.php?topic=7435.0


Mike

[David]

Hola Andrés.
Magnifico trabajo como siempre!

(Permitidme escribir en castellano, traducir todo esto en google puede ser peor que no hacerlo)

He seguido tu labor a raiz de las fotografias de gran campo algo mas "artisticas" que he podido recopilar este verano, donde incluyo una buena parte de suelo.
Le he dado mil vueltas a la manera en la que podria hacerlo con PixInsight, pero ni de cerca he podido hacer nada. Ya desde el principio del procedimiento que te he leido me daba error, no he sido capaz ni de resolver el campo... luego ya no he seguido por que dudaba que fuera a salir bien las uniones del suelo.
Estoy saliendo del paso con programas alternativos, que me lo une decentemente, pero me interesa poder hacerlo con PI por muchos motivos, uno de ellos es probar con imagenes lineales o alteradas lo menos posible, y otro intentar evitar gradientes creados en otros programas.

Aqui abrí un hilo sobre mi cuestión:

http://pixinsight.com/forum/index.php?topic=7134.msg48163#msg48163

Tengo unos cuantos ejemplos mas recientes,
6X3:
https://www.flickr.com/photos/david_astrofotografia/14619716796/

3X3:
https://www.flickr.com/photos/david_astrofotografia/15038775481/


Perdona que abuse asi de tu hilo, pero el sistema que propones me parece tremendamente bueno, si pudiera aprovecharlo para este tipo de foto podria ser fantastico el resultado.
Crees que es viable conseguir algo en este tipo de foto con tu metodo?

Saludos, gracias, y perdon de nuevo por el abuso.

[georg.viehoever]

That is very impressive. I cant wait to use it on a couple of my widefields.
Georg

[cs_pixinsight]

Andres, has this new version of the ImageSolver been released at this point?

I'm trying to use v3.51 on Canon 6D images using a 50mm lens and it just doesn't want to solve to a high enough accuracy.

[Andres.Pozo]

Hi,

I have a new version mostly finished since the beginning of September that requires some new capacities available only in PI 1.8.3. The problem is that 1.8.3 wasn't published until much later and I am now extremely busy in my job.  My plans are to finish it during my Christmas holidays but I can't promise anything.

[cs_pixinsight]

Andres, I can understand the work demands.  Thanks for the update and I hope you have restful holiday season.

Craig