No, perhaps I didn't make the situation clear enough. My observing site has pretty severe seeing, compared to pristine observatory sites. My seeing disk is typically about 7 arcsec. My pixels are individually about 1.8 arcsec. But they are arranged in a Bayer matrix. Hence, with these conditions, SuperPixel deBayering produces probably the very best color distribution that one could hope for.
Drizzle is used to slightly increase the resolution of your image data by taking advantage of randomly sampled sub-pixel phasings, and drizzling each sample onto a higher density sampling grid. In this default case, we are upsampling by 2x and dropping our pixel data as though they were 0.9 pixels in size. The resulting drizzle map shows pretty good coverage, worst (around 0.6) at the edges of the stack, where you'd expect the failure of overlap due to dithering.
By using SuperPixel deBayering, I'm routinely seeing good star images, with no undue color fringing. I'm using a mostly reflector telescope with only a few field lenses in the HyperStar assembly, and that front corrector plate on the C8 of course. So I wouldn't expect the optics to produce the kind of color fringing that is more common with camera lenses and refractors.
There have been some claims made that Drizzle integration produces better, more round, star images. In looking at my own data where I compare a VNG luminance image against a Drizzled SuperPixel luminance image, using the FWHMEccentricity tool on a very busy star field, like around the Eastern Veil Nebula, I find that claim to be false. The VNG image shows better statistics, but wider variation across the FOV. The difference between them is slight, however.
There will be some degree of false increase in high spatial frequency resolution when using any kind of interpolation onto a higher density sampling grid, as happens with VNG deBayering. It is very easy to see that with the FFT magnitude of an image. The Bayering matrix also produces its own aliasing patterns in those FFT images, different for B & R, versus the pair of G channel pixels.
And drizzling will likewise produce some degree of high spatial frequency enhancement, but of a different kind. My VNG images show significantly more power in the higher spatial frequencies than the SuperPixel drizzled images. And I might expect that. But anything produced at those high spatial frequencies in the VNG image, where those spatial frequency regions overlap the aliasing islands produced by the Bayer matrix, are suspect. By contrast, the drizzled SuperPixel images seem to respect those aliasing regions and avoid placing much power in them.
... so I forgot what I was asking about.... perhaps I just answered my own question...
