Both frames are stacks of 5 x 150s. The left panel shows the ATIK 490 in 1x1 pixel binning mode (full color OSC). The right image shows 2x2 pixel binnning mode, where all 4 pixels in every Bayer cell are simply co-added.

Both images were stretched equally in luminosity with PixInsight. The color image was additionally stretched in chrominance. No attempt was made to diminish the star contribution. Background and stars were stretched equally together.

It would seem reasonable that 2x2 binning would be more sensitive to extended regions, since (A) only 1 readout noise per Bayer cell, and (B) the pixels are effectively larger by a 4:1 area ratio. And the results seem to bear out those facts if you look at the very faintest background cloud regions.

But despite those facts, I can't help but feel that the full color image is still showing more information about the region. The dimension of color adds quite a bit. And even in the faintest cloudy regions, the color difference with its background makes it appear to stand out better than in the super-pixel version. It may well have lower SNR in those regions, but the color image still grabs my attention better.

I suspect too, that simply adding the 4 pixels from every Bayer cell together will not be a quadrupling of sensitivity, as you might expect from area ratios. But because the 4 contributing pixels don't have the same spectral response, the summed result will probably be about half or so, compared to adding 4 identically responding pixels together.

What do you think?

-----------

The second pair shows the results after much creative stretching of both the color and the B&W images. In the color image I used a local histogram equalization to bring out more detail in the nebula. In the B&W image I enhanced the background by adding a stretched multiresolution support layer at long scale lengths to avoid stars.

The results are quite interesting. Remembering that the B&W image was produced by running the camera in 2x2 binning mode, we have less resolution to work with. But the background shows up much better after stretching than in the color image. Meanwhile the color image beats out the B&W with fine detail in the nebulosity.

But in the end, it really does appear that the 2x2 binning mode is more sensitive to extended regions. Look at the extended clouds surrounding the central nebulosity in the B&W image. The color image shows just the barest hint of those wider cloudy regions.

It just takes a bit of stretching work to bring out the low intensity information.

------------------

So... what if we take the processed images and construct an LRGB 4-color image? We take the stretched and detailed color image, downsample by 2x2 to create a matching frame size for the B&W image, and use its R,G,B components in combination with the B&W image as the L channel?

That's what we get in the final revision of the image. This is a bit backwards from the usual approach of taking a high resolution L image in combination with lower resolution color information to construct an LRGB image. But this is what we have at hand, so why not?

------------------

And so here we are at version 4, where I applied nothing more than a strong chrominance stretch and a local histogram equalization.

Is this just junk? or has the 2x2 mono image actually provided more information along with the low grade color information in the surrounding clouds? The B&W provides intensity information, and the color image just provides coloration for each region. There wasn't enough intensity in the surrounding clouds in the color image to provide anything near to this level.