NB From Colour

[Brian28]

 

I read somewhere on this forum a comment that you can take colour images and convert to narrowband ?  And then process .. or have I got the wrong end of the stick ? Was it just a Luminance frame ?

can someone put me right please ? ?

 

cheers Brian 

 

[vlaiv]

In theory, if your target is composed only of certain narrow band wavelengths (like emission nebula containing Ha and OIII) then you can theoretically split those into two monochrome NB images (red channel would be Ha, from blue and green you can extract OIII). If target has Ha and SII in red channel - there is no theoretical way to distinguish between them - camera does not record wavelength information of particular photons.

Easier thing to do is to extract Luminance. Well, not really that easy if you want to do it properly, but it can be done for any RGB/OSC image. Simplest method (and not always accurate) is to sum R, G and B. This is also the worst method in terms of resulting SNR. There are other methods like converting image to Lab or other color space that has Luminance as component and using that. You can also use weighted channel addition - this is in essence how L is produced in mentioned color models (specific weights), but you can assign your on weights, depending on data.

If you want to get the best Lum from RGB in terms of SNR, that is not an easy task - each R, G and B will have different SNRs and just simply adding them will not yield best results, neither will assigning global weights. You need to combine them with adaptive per pixel weights that preserve total signal ratios across the image but minimize noise contributions from each channel. Just to make it clear what I'm taking about, imagine zone in image with pure red signal - like Ha nebula. SNR in red channel will be good, but in green and blue it will essentially be 0 since there is no signal, only background noise. If you add these three channels you will be left with only R signal, but R, G and B noise combined - that is the same signal and more noise - worse SNR. In that particular case you are better off using R channel only then combining them (in that image zone). Imagine now different case (almost opposite one) - R, G and B channels contain same amount of signal (some sort of reflection nebula that is greyish in appearance). Using addition / average in this case will be best approach, because it will behave like stacking and will improve Lum SNR by factor of 1.732 (square root of 3). So you can see why this can be complicated subject - you can have many different parts of image having different ratios of signal per channel (different colors) and you would need to find best solution for each part of image to maximize SNR gain of creating Lum.

Another simple "fix" that will not improve SNR but will not degrade it either is to use Max of (R, G and B ) as Lum, or signal with max snr for particular pixel - people often do this without knowing when processing their NB images - they for example use Ha signal (often the strongest and therefore best SNR) as Lum layer when doing false color images.

[Demonperformer]

There has been chat about using this filter to capture Ha & Oiii on a colour camera at the same time.

[Brian28]

Wow thanks Vlaiv ..   ? I will read that 3 or 4 times to get my head round it ..  

 

and thanks DemonP .. I wonder if that filter would work with my IDAS D1 , to increase the Ha ? 

 

Thanks Guys