RE: [APML] ESO Tarantula CCD image

[Jerry Lodriguss <jml@astropix.com>]

Hi Chris, Sean,

I stand corrected. You are very right!

I can't say I have a clue as to why this works, but I did a bunch of 
experiments last night and the numbers back yall up.

I did find out a couple of interesting things. Just converting the color 
image to grayscale gives a better signal to noise ratio than averaging the 
individual color channels into a new luminance image.

At least, the way I averaged them, it did. But, maybe I didn't average them 
the right way.

Anyway, I have all of the results of my experiments at:

http://www.astropix.com/HTML/TEST/TEST1.HTM

It's pretty graphics heavy, so it may take a while to load for those with 
slow connections.

I also have the original color image available for download from that page 
as an uncompressed TIFF if anyone would like to verify my findings, or try 
new experiments.

I learned something new last night, and it was fun. So, thanks Chris and Sean!

Jerry


At 02:25 PM 6/14/2002 -0700, you wrote:
>Jerry, my understanding on how this works is that in an image, the grain is
>essentially a luminance noise, and much less color data noise.  If I add the
>R+G+B and say thats my new luminance, it will of course have much less noise
>than any single frame.  That appears to be the secret, a standard RGB
>combines the three channels in such a way that the noise is "colored" and
>much more visible.  Some people even blur the RGB and combine with a R+G+B
>luminance to make an even smoother but equally sharp image.  This is the
>best I can explain it.  I have used it dozens of times, and done carful
>comparisons.  It works.  You might try splitting an RGB 16 bit image in PS,
>and make a Lab this way to compare.


Astronomical photography: http://www.astropix.com

Sports Photography: http://www.astropix.com/SPORTSPIX/INDEX.HTM


--  APML Archives at <http://astro.umsystem.edu/apml/>  ---
             Unsubscribe at <majordomo@seds.org>