Lunar/Planetary Alignment

[Enzo De Bernardini]

Hello,

I have taken a sequence of photos of the last lunar occultation of Saturn using only a camera and tripod, but it has been impossible to align the images (which have displacement, some rotation, and other typical atmospheric effects). I have tried with no success FFTRegistration script (successful on a large scale, but visibly not accurate) and StarAlignment (also unsuccessfully in my case).

My wish: A la PixInsight dedicated tool for lunar and planetary registration (drizzle support?), with batch alignment using a reference frame (and maybe a quality estimator, for large planetary image sets for use only the n best).

That's it

Greetings,

Enzo.

[Carlos Milovic]

Hola Enzo. Si tienes Linux, prueba el módulo TaylorRegister que está en mi sitio. No alinea distorciones locales (tal como está hecho ahora), pero puede que sirva para la alineación global.

Saludos,

[Enzo De Bernardini]

¡Gracias Carlos!. No estoy usando Linux, pero veré de pedir asistencia a Alejandro para la tarea y probar el módulo, a ver que tal funciona con las imágenes.

Saludos,

Enzo.

[Carlos Milovic]

Ok! Cualquier cosa, que me mande un mensaje.

[Alejandro Tombolini]

Hi!!!

I just try with a set of Saturn images, previously debayered and work pretty good, but the main problem I think is the classification of the images. We will need something like sub frame selector but for planetary images! It really will be great.
For example in this set 5000 images, I took 100 and the result with borders not defined. Inspecting the images with blink, only 10 seems good, the rest are very bad.
Saludos, Alejandro.

[Juan Conejero]

Come on guys, don't hesitate to ask for the Moon 

Seriously, we have so many and so complex pending tasks, that the planetary processing engine that you are asking for is unlikely to happen. A little example so you can figure out: with 4K and Retina monitors becoming standard in a few months, I have to redesign the whole graphical interface of the PixInsight Core application, including different versions of every icon and a new control rendering engine. Add to this the documentation, new scripting engines, GPU acceleration...

[Carlos Milovic]

If it doesn't take much time, I may write a "dirty" module/script that makes the classification. I would need some help in the research, finding some metrics or methods to implement.

[vicent_peris]

Hi,

We have talked about planetary imaging many times, but I realize this is a complex topic and I don't see a good solution implemented in PixInsight at short or mid term. But, IHMO, at some point in the future, we'll decide to go for it. I do think we should review all these techniques, as I particularly don't like the existing solutions available today. I see a point of inspiration in some techniques applied to medical imaging, which are designed to register deformed images.


Regards,
Vicent.

[Carlos Milovic]

Vicent, do you remember Michael (a co-worker here)? He was working on motion vectors, that were pixel-based. It is a possible approach, very flexible, used a lot in video super-resolution. This might be the next step. But first, grading

[Enzo De Bernardini]

I know, I know... Next time I will capture the images with the EQ mount for no field rotation...

[Carlos Milovic]

Enzo, if TaylorRegister does its job, it should handle rotations. It is not a very robust algorithm, but should work fine with high SNR data.

[Enzo De Bernardini]

Thanks Carlos!

(chances to compile for Win 64?)

Greetings,

Enzo.

[Carlos Milovic]

I tried, but we have a compiler error in windows only. It seems to be triggered in the PCL, so Juan is looking at it.