A couple of doubts about ColorCalibration

[Jordi Gallego]

Hi Juan and everybody,   (castellano abajo)

I have a couple of doubts and/or questions regarding ColorCalibration tool.

First, how should one interpret that the white reference mask is a color image instead of a grayscale image? I do not understand how such a mask works. Can you please explain this?

Second, (this is maybe more a request than a question), why the color calibration factors, Kr, Kg, Kb, that are displayed on the processing console after the application of the process, are higher than one?. I know that is a current practice on other software packages but this is, I believe, against PixInsight philosophy. Is it possible to change these coefficients to the normalized range 0 -1?, or alternatively, if this is not possible or desirable, is it possible then to change the presentation of the manual color correction sliders to accept numbers higher than one?

Thanks in advance
Jordi

Hola
Tengo un par de dudas y/o preguntas acerca de la herramienta ColorCalibration.

En primer lugar, ¿cómo se debe interpretar que la máscara white reference sea una imagen de color en lugar de escala de grises?. No comprendo como trabaja una máscara así. Juan, puedes explicarlo por favor?

En segundo lugar (ésto es más una petición que una pregunta), ¿Porque los factores de calibación de color que aparecen en la processing console tras la aplicación del proceso, son mayores de uno?. Ya sé que esto es práctica habitual en otras aplicaiones, pero entiendo que esto va en contra de la filosofía de PixInsight. ¿Es posible cambiar la presentación de estos coeficientes para que queden dentro del rango normalizado 0-1?, o alternativamente, si esto no es posible o deseable, se puede cambier la presentación de los manual color correction sliders, para aceptar números mayores de uno?

Gracias anticipadas
Jordi

[C. Sonnenstein]

ColorCalibration is not a maskable process. In other words, ColorCalibration doesn't use pixel values to perform a calibration process with the white reference image. The calibration task is done with the global statistics around the white reference image. When you know the physic propierties of the object you can tell to ColorCalibration process the correction factors. Only if you enable the output white reference mask, the white reference mask will be white for pixels have been used to perform color correction factors.

Cheers,

[Jordi Gallego]

Thank you for your fast answer Carlos (it is really funny to talk with you in english )

Carlos, I probabily did not explain well my first question: I know that ColorCalibration is not a maskeable process, but what I intended to explain is that when you enable the white reference mask, you get a color mask. And I do not know how to interpret the action of such mask: for example what is the meaning of a red pixel on that mask, this pixel for example has been used only to compute red channel???

Thanks and regards
Jordi

[Niall Saunders]

Jordi,

What if you try to ColourCalibrate a 'known' image.

What about somw sort of 'test card' - if you don't have one, or cannot create one, I can send you the one that I used to help me write my deBayer PJSR script.

Basically it has six squares one it, one for each primary (RGB) colour, and one for each secondary (CyMgYe). It also has areas of 100% 'grey' (i.e. pure white), 75% grey, 50% grey, 25% grey and 0% grey (i.e. pure black)

That kind of image might at least let you interpret the 'white mask' better.

You should also be able to take the 'pure' image, and throw a nasty colour (hue) curve on it, amtbe even a funny Histo, and then see how, or if, colour calibration and background neutralisation can recover - give that you can define what is supposed to be 'background' and what is supposed to average out as 'white'.

Just a thought - I don't have time to try this myself - but I could try and find my test image, and upload that as a FITS file, if you think it would help.

Cheers,

[Jordi Gallego]

Thank you very much for your interest and your answer Niall!

If it is not a big trouble for you, yes please, I would like to take a look and do some tests on your test card

Thanks again
Jordi

[Niall Saunders]

No problems - I will dig it out and post it tonight, if not here on the Forum, then 'directly', by email.

Cheers,

[Niall Saunders]

Hi Jordi,

OK, here is a 'brand new' colour chart (almost identical to my original, but I decided to rebuild this one with absolutely NO possibility of ICC profiles getting in the way)

The original is made in CorelDraw (by far the easiest and quickest way to do this sort of thing !!) and I have made available both a 640x480 BMP and an RGB FITS version - it doesn't matter which you use, once they are in PI they are identical - obviously the FITS version is a much larger download (so just follow the links)



And the download links are ( use <Right-Click><Save Targt As ...>, or the equivalent for your Browser)

BMP   (900Kb) http://pixinsight.com/nsaunders/NewSimpleTestCard.bmp
FITS (3.51Mb) http://pixinsight.com/nsaunders/NewSimpleTestCard.fts

Have fun,

[avastro]

Hi Niall,
Thanks for your test card color.
The FITS link is also directed to the BMP image.

Cheers

Antoine

[Niall Saunders]

Hi Antoine - thanks for pointing that out.

Updated, and file sizes added - and dowloads checked this time !!

Cheers,

[Jordi Gallego]

Thank you very much Niall

Jordi