LRGB vs CRGB

LRGB vs CRGB

LRGB vs. CRGB Discusión del bloqueo de luminancia con ejemplo de NGC2403

Don Goldman

PLANTEAMIENTO DEL PROBLEMA

El uso de un filtro de luminancia con bloqueo de infrarrojo o de un filtro sin bloqueo (CLEAR), para porción de Luminancia de alta resolución de una imagen ha sido y sigue siendo motivo de un debate acalorado en la comunidad de astro-fotógrafos, Hay muchas opiniones, conjeturas y creencias, pero pocos datos. El trabajo que sigue en un intento sistemático de recoger datos para diferenciar que puede ser ficción o realidad.  

 

La idea del método LRGB del Dr. Okano es dejar a la imagen de la luminancia proveer de detalle y dejar el RGB, o crominancia, para que le dé color a la luminancia. Ésta es la razon por la que tomamos Luminancias sin hacer binning a la máxima resolución de nuestra CCD y tomar el RGB en binning 2x2, a menor resolución, para una fotografía "más rápida" y muchas veces "desenfocar" suavemente la crominancia. Combinamos entonces la luminancia de alta resolución y la crominancia RGB en capas separadas en programas tales como Adobe Photoshop, dando como resultado una imagen en color de alta resolución. La pregunta es: ¿Cuál es la mejor porción de luminancia?. ¿La que tiene bloqueo de infrarrojo o sin bloqueo o CLEAR?. 

Hay dos escuelas de pensamiento en éste asunto, que actualmente sigue un debate caluroso entre la comunidad de astrofotógrafos:
  1. Elegir un filtro de luminancia L con bloqueo de infrarrojo NIR-blocked que se adapte a los filtros RGB para una mejor integración de color, pero sacrificando un 25% (STL11000, ST2000XM)  o 45% (ST10XME) de la señal de luminancia proviniente de la señal cercana al infrarrojo.  Éste planteamiento asume que la señal adicional del cercano al infrarrojo arriba mostrada, no es "señal pura" pero contribuye con ruido a la imagen LRGB porqe no hay fotones de RGB que directamente coincidan con los fotones del NIR (near infrared).  Añadiendo más "ruido" a la imagen no mejora su calidad.
  2. Elegir un filtro sin bloqueo(tipo C : Clear) para una mejor señal de respuesta y conseguir todos los fotones posibles en el tiempo de apilado. Cualquier falta de coincidencia con los datos de color es ligera y puede ser procesada.

DATOS

Estos datos fueron recopilados por Ken Crawford en su observatorio Rancho del Sol observatory a 3000 pies en Sierra Nevada foothills del norte de California con telescopio de 20 pulgadas RCOS Ritchey-Cretien, una cámara SBIG ST10XME CCD y filtros Astrodon® Tru-Balance LCRGB. 60 minutos de L y C fueron tomados sin binning y 42 minutos de cada R, G y B en binning 2x2. Se procesó el RGB.  La imágenes de Luminancia (L) y Luminancia con Clear (C) fueron procesadas lo más identicas posibles. Se componieron imágenes LRGB y CRGB en Photoshop CS, tan idénticas como fué posible.

DISCUSIÓN

Desde que hago fotografía, ha existido debate sobre si usar un filtro de bloqueo para el cercano al infrarrojo para la porción de alta resolución de nuestros datos LRGB, o un filtro clear que no tiene ningún bloqueo al ultravioleta (UV) o cercano al infrarrojo (NIR) blocking. Sólo tiene revestiemiento anti-reflectante. La comunidad de SBIG Yahoo Group se mostró uniformemente dividida. Como en muchos asuntos de nuestra afición, hay muchas opiniones, rumores y espectaciones, pero no mucho en datos puros y duros que ayuden a crear una opinión informada. 

Antes de continuar, estableceré una terminología consistente. Propongo que utilicemos Luminancia (L) para hablar de un filtro con bloquo NIR y (C) para uno sin bloqueo.

Tengo mi propia perspectiva en éste asunto. He concluido hace algunos años ya que creo que los filtros L, al ajustarse a los parámetros de los RGB proveerán el mejor color y detalle, especialmente en galaxias. I have my own perspective on this issue.  I have stated for several years now that I believe L filters, matching the extents of the RGB filters will provide the best overall color and detail, especially for galaxies.  The C filter provides a 45% improvement in signal from NIR photons for an ST10XME and about 24% for an STL11000.  Since, NIR photons don't have direct analogs within the RGB space, the extra signal could be considered "noise", akin to sky glow.  Further, the signal will mostly affect reddish objects, so the 45% improvement will not be seen "across the board". 

Note.  I recognize that the NIR photons are just part of the blackbody curve generated by stars in the galaxy that are present in the RGB filters, with cooler stars contributing relatively more. So, there is some relationship.

I tried to show some of this with Hubble's Variable Nebula, where, with the exception of some fatter red stars, there was little or no improvement in the nebula using a C filter over an L filter.   Click here for the link to that report.

The limitation of that study was that is was done on a bluish object rather than on a galaxy having a yellowish core, bright HII regions and extended blue spirals.  Fortunately, my close imaging friend, Ken Crawford, took an image of NGC2403 in Camalopardalis in December and took RGB, L and C data.  His finished image combined both L and C into the "luminance" layer.  Ken provided me with his master L, C, R, G,  and B FITs files. These were all taken on the same evening. My goal was to process the RGB TIF file and then as best I can process the L and C files similarly, and make LRGB and CRGB composites.  The results are shown below.

The immediate overall impression in comparing the LRGB and CRBG images above is that the CRGB galaxy seems to have a "glow" around the galaxy that washes out the color and appears to diminish contrast.  The opacities of the L or C layer in Photoshop CS (luminosity blending mode) were set to 60%.

The next comparison is of the two luminance images using MaximDL's information window. Identical rectangles were drawn at the same place in each image; a place in the core and another within the background without including any stars.  The average ADU for L and C in the core was 508 and 829, respectively.  I have shown in the Hubble's Variable Nebula report that the Standard Deviation  (Std. Dev.) value provided in the Maxim Information window is a good approximation of signl-to-noise (S/N).  The std. dev. values for these core samples are 98 and 166, respectively.  So it appears that the C filter delivers much better S/N.

MaximDL Analysis of S/N in the Luminance (left) and Clear (right) Filter Images

 

The background areas were also analyzed similarly within identical rectangles near the upper right portion of each image. The average ADU values for L and C are  216 and 379 and std. dev. values of 10 and 15, respectively. 

So, the C filter provides the predicted (from the quantum efficiency [QE] curve of the CCD detector [KAF3200ME]) improvement of of ~1.5 over the L filter. It provides a higher S/N, albeit over a brighter background.  This creates the "glow" described above as can be clearly seen in these "luminance" images. 

Comparison of Equally Stretched Luminance (left) and Clear (right) in Photoshop

The greater S/N from the C filter sits atop a brighter sky background, or "galaxy glow" due to the extra NIR photons.  To me this is like imaging the Horsehead Nebula with RGB filters from a large city.  Fainter detail will tend to be washed out.  This is not so much a problem with a bright galactic core.  It will be noticed more as a loss of detail in fainter regions, e.g. between the spiral arms. I think this effect can be seen above.

The plate scale of Ken's 20" RCOS RC with an Astro-Physics' 0.67x focal reducer is 0.44"/pixel in the ST10XME.  Using that setting in MaximDL, a number of stars in each "luminance" image were analyzed with Maxim's Information Window. Those data are summarized in the table above.  There is a systematic increase in the full-width-at-half-maximum (FWHM) of the stars from the C filter in comparison to the L filter. Overall, there was a 0.08" increase in star size in the C image.  Of course seeing could change within the course of imaging throughout the evening, but there are reasons to expect this, as discussed next.

Analysis of Full-Width at Half Maximum Intensity (FWHM) for Stars

It could be argued that the focal reducer is the cause.  However, in discussing this with Roland Christen at Astro-Physics (personal communication), he mentioned that the Airy disk in the NIR (e.g at 1000 nm) is nearly twice that in the visible.  So, even if we had perfect optics, there would be a difference.  The C filter would degrade contrast. Refractive optics will just make it worse. This is the reason why L filters are strongly recommended for people using telephoto lenses on digital cameras (e.g. Canon 20D) to minimize this refractive optics dispersion problem. 

RESUMEN

If your goal is to bring out magnitude 22 galaxies by combining many long "luminance" exposures, then it may be best to use a C filter or no filter at all.

If your goal is to bring out vibrant color in galaxies, these tests suggest that the L filter, matched spectrally to the RGB filters, is a better choice to eliminate the NIR "galaxy glow" contributed by a C filter.   This becomes even more important in systems with refractive optics (SCTs, refractors).

If you have a refractor or telephoto lens, remember that a C filter also lets in UV. This is why less color-corrected refractors often need "minus violet" filters to eliminate the optical dispersion in the short wavelength region (blue halos around bright stars). An L filter does cut out some of this. For these systems, an L filter is likely to be a better choice. 

AGRADECIMIENTOS

Thanks to Ken Crawford for making his data available and Roland Christen for discussion of the optics.