Adding RCD debayer method

[minhlead]

I am using Syril alongside with PI and since 0.99.4 Syril has added the RCD debayering method which significantly improves the resolution and reduce artifact around dim stars (some how my stars are getting colors dots around them when debayer with VNG)
Could you please consider adding RCD as a debayering method for PI?
@Juan Conejero

 

[Juan Conejero]

Thank you for this suggestion. I assume you refer to this:

GitHub - LuisSR/RCD-Demosaicing: A Bayer demosaicing routine aimed to reduce pixel overshooting

A Bayer demosaicing routine aimed to reduce pixel overshooting - LuisSR/RCD-Demosaicing

github.com

It looks interesting and we'll try to implement it as soon as possible in our Debayer tool.

Anyway, the only optimal way to integrate CFA data is drizzle integration, simply because drizzle avoids interpolation completely. No interpolation method can be better than no interpolation at all. For this reason, all debayering tasks generate just temporary/disposable data that should only be used to update drizzle files (.xdrz) with the StarAlignment and ImageIntegration tools. The actual integration of CFA frames should always be performed with the DrizzleIntegration tool, with the Enable CFA drizzle option checked, and a drizzle scale of 1, or higher if you really want to use drizzle to reconstruct undersampled data at the same time.

 

[minhlead]

Thank you for this suggestion. I assume you refer to this:

GitHub - LuisSR/RCD-Demosaicing: A Bayer demosaicing routine aimed to reduce pixel overshooting

A Bayer demosaicing routine aimed to reduce pixel overshooting - LuisSR/RCD-Demosaicing

github.com

It looks interesting and we'll try to implement it as soon as possible in our Debayer tool.

Anyway, the only optimal way to integrate CFA data is drizzle integration, simply because drizzle avoids interpolation completely. No interpolation method can be better than no interpolation at all. For this reason, all debayering tasks generate just temporary/disposable data that should only be used to update drizzle files (.xdrz) with the StarAlignment and ImageIntegration tools. The actual integration of CFA frames should always be performed with the DrizzleIntegration tool, with the Enable CFA drizzle option checked, and a drizzle scale of 1, or higher if you really want to use drizzle to reconstruct undersampled data at the same time.

Thank you for this suggestion. I assume you refer to this:

GitHub - LuisSR/RCD-Demosaicing: A Bayer demosaicing routine aimed to reduce pixel overshooting

A Bayer demosaicing routine aimed to reduce pixel overshooting - LuisSR/RCD-Demosaicing

github.com

It looks interesting and we'll try to implement it as soon as possible in our Debayer tool.

Anyway, the only optimal way to integrate CFA data is drizzle integration, simply because drizzle avoids interpolation completely. No interpolation method can be better than no interpolation at all. For this reason, all debayering tasks generate just temporary/disposable data that should only be used to update drizzle files (.xdrz) with the StarAlignment and ImageIntegration tools. The actual integration of CFA frames should always be performed with the DrizzleIntegration tool, with the Enable CFA drizzle option checked, and a drizzle scale of 1, or higher if you really want to use drizzle to reconstruct undersampled data at the same time.

Thanks you. That's great to hear.
I understand that bayer drizzle can give excellent result but I tends to avoid bayer drizzle since under certain circumstances, it can introduce some undesirable noise into the intergrated. RCD give somewhat noisier background than VNG but excellent saturation and significant more resolution especially in stars edges.

 

[Juan Conejero]

I understand that bayer drizzle can give excellent result but I tends to avoid bayer drizzle since under certain circumstances, it can introduce some undesirable noise into the intergrated.

Not at all if you acquire a sufficient number of well dithered frames and restrict drizzle to unit scale. Both conditions (enough frames and good dithering) are absolute requirements anyway, irrespective of how you are going to integrate your data.

 

[minhlead]

Not at all if you acquire a sufficient number of well dithered frames and restrict drizzle to unit scale. Both conditions (enough frames and good dithering) are absolute requirements anyway, irrespective of how you are going to integrate your data.

Well I do not dither since my mount have ample of DEC backlash and it never settles after dither. So that is the reason I am seeing an increased of noise by doing bayer drizzle may be?
Also almost non artifact on star edge with RCD is a big plus for me since I usually push the saturation of the stars up quite a bit and VNG debayer seems to put color artifact into the star edge sometimes and they can be annoying when the saturation is increased.