Mar 26, 2016 9:58 AM
| Last Modified: Feb 10, 2017 5:41 AM
Go somewhere away from the city lights. Put an inexpensive camera on a tripod, point it at the night sky, fiddle with the settings, wrap a rubber band around the button to make it take pictures for an hour. The pictures are really dim, but with some simple JSL processing to remove some noise and increase the contrast...
Yes, it was really, really dark, and impossible to tell that fence was in the frame. It won't matter, much.
Next, stack the pictures. This requires finding a transform matrix for each image to warp it onto the base image, and adding up the values. Here's the Pleiades cluster from the center of the video, enhanced by combining all of the images...
After writing all that JSL, why are there black corners around Alcyone? ( al-sy-ə-nee according to Wikipedia ) The answer lies mostly in the camera, partly in the JSL. Here's a 16x16 pixel snip of Alcyone...
That's not a black square, but it is close. This is from one of the JPG images from the camera. Here's the gray scale matrix of pixel data, R+G+B on a scale of 0 to 255+255+255:
The original image from the camera has dark corners too! A serious amateur astronomer would get a better camera, with RAW picture output (and a better lens and a better sensor and a more stable tripod...). This camera has almost certainly applied a bi-cubic sharpening filter to the captured data before creating the JPG image. A cubic interpolation creates the over-shoot.;Bicubic interpolation - Wikipedia, the free encyclopedia As far as I can tell, I didn't ask for that.
The JSL sees those 3, 5, 8 values (sum of r+g+b) and subtracts the noise floor, which is typically 4-6 for each of the r,g,b channels and winds up with a negative brightness, which is clipped to zero.
Turns out there is a lot of stuff I don't know much about...
update: tried subtracting median to remove noise. Mean is too big and makes a black streak if a pixel ever has a star pass over (see Alcyone above). Min is too small and produces a light streak. Not sure where the cloudy edge is coming from, but the center looks better:
I'll try to get all that JSL organized for another post soon. For now, the 16x16 pixel contour plot JSL is attached.
update 10 Feb 2017: repair formatting for new community