Imaging Articles
Does Binning Color data 2x2 Matter in LRGB Composites?
There has been considerable discussion over the years that CCD tri-color imaging has been practiced by amateurs about whether
it's best to do the RGB color frames of an LRGB composite at full resolution (1x1) or binned 2x2. Binning the color frames provides
a more sensitive imaging response from the CCD chip, allowing the imager to make shorter exposures through the color filters. There are still
a number of imagers, however, who insist that shooting the color frames at full resolution (1x1 binning) provides a higher-resolution final LRGB image.
Being generally the lazy sort, the idea of shorter exposure times for the color frames by binning them 2x2 appealed to me, but I didn't want to sacrifice
image quality for the sake of my laziness! So, I decided to see for myself how much binning color frames affected image quality in an LRGB image.
In 2002 when I did an image of M16 from my backyard, through a Meade 6" Schmidt-Newtonian at f/5, I shot 80 minutes of luminance data at full resolution through
the normal red filter. I then did 2 sets of color data files: the first set used 2x2 binning for the RGB data (40:40:40 minutes done as 20x2 minutes each),
and the second set was shot at full resolution (1x1 binning, 80:80:80 minutes done as 20x4 minutes). I had wound up using the binned data for my posted image, and
at the time looked at the full-res data to see if there was any difference, but had never done a full-blown visual comparison until now (Sept. 2003). Here are the
results of the comparison.
First, there are three images, all RGB composites. The first is the 2x2 binned image, the second is the 1x1 binned image, and the third is the 1x1 binned image to
which a 4.0 pixel Gaussian blur has been applied in PhotoShop. I had found that a slight blur reduces color noise in final images before, but wanted to see how much
the blur affects final image resolution. Note: all the images below are shown at 1/2 full size to save space!
Color data binned 2x2
Color data binned 1x1 (full res)
Color data binned 1x1 (full res), 4.0 pixel Gaussian Blur
I used levels & curves in PhotoShop to match the brightness, color balance, and saturation of the two main images (binned 1x1 and 2x2) as closely as possible for the
fairest comparison. In reality, the 1x1 binned data is not quite as *deep* as the 2x2 (since to match exactly, exposure time should have been 4X the 2x2 data, but
in reality it was only 2X), yet the 2x2 data has a bit more blue tint that couldn't be removed (the moon was up while it was being shot!), and had more
trailing because my mount at the time was crossing the meridian! All in all, though, they're pretty close.
Next is the luminance image (shot through a red filter, 80 minutes as 20x4 minute exposures). The exact same luminance image was used to build LRGB images using the three
RGB color files above -- no changes at all to the luminance file in resolution, contrast, brightness, stretching, etc. was done between building composites. Here's the luminance image:
Luminance image, 80 min. through red filter at full resolution (1x1 binned)
Here are the three LRGB images that resulted from making the three seperate LRGB images -- note that I'm only showing a small portion of the center of each image, but at full resolution (not resized)
for space reasons. Each LRGB composite was done in PhotoShop, with the luminance opacity and fill both set to 100%.
 |
 |
 |
LRGB using 2x2 binned RGB data
|
LRGB using full-res RGB data
|
LRGB using full-res RGB data Gaussian blurred
|
Here are my conclusions from this exercise...
I can notice what appears to be a bit rounder stars in the extreme upper-left corner of the 2x2 color image (far left) -- you can see a bit less trailing in the original 2x2 binned RGB image
if you zoom in on it, due no doubt to the shorter 2-minute exposures, and this came through a little bit in the final image. The 2x2 version is a bit less saturated in red and shows a slight blue
tint compared to the 1x1 image, which is mostly due to my not-entirely-successful attempt to perfectly color balance the RGB files more than anything else. Actual resolution, sharpness, and color
transition information in the two images is as identical as I can tell from examining them. I can see no improvement in the full-res RGB image that clearly shows greater resolution or color fidelity.
The interesting thing to me is the LRGB image made from the blurred data -- the blurring was pretty strong, yet shows very subtle effects in the final image. There is certainly a more visible "softness"
to the LRGB made from the blurred data, but to my eye it's an appealing softness that smooths color transitions and reduces color noise without affecting overall sharpness of the image. To my eye, the
blurred version is the most appealing of the three images, but you can draw your own conclusions.
Others have pointed out very valid reasons for shooting the RGB frames at full resolution: using some of the RGB frames mixed in with the luminance data to increase the S/N of the luminance frame,
shooting at the same binning to use the same dark frames, avoiding gross undersampling, etc. Leaving these mitigating reasons aside, however, I can see no real advantage to the longer exposure times required to shoot color at
full resolution, and will continue my lazy ways of binning color data 2x2 to save time and to gather more color data in the limited dark sky time I do get!
Back to the Articles Index
All images Copyright (c) 2002-2003, Paul LeFevre
Mail me with comments & criticisms!