One of the attraction of this Cable and Igor's (CAI) adventures series is the ability to make sense of what is traditionally considered "noise". After, the radiation deconvolution example (more to come on that later), Cable performed a Robust PCA (using GoDec but any other similar algorithm should do as well) of the craters of the moon as captured through a telescope.
The person who did this very nice shot, Paul, goes on to explain:
"...Live view of craters on the Moon through my telescope using a Philips SPC900 VGA CCD webcam. The shimmering is caused by atmospheric turbulence and is the same phenomenon that causes the stars to twinkle. This video nicely demonstrates why the atmosphere is the limiting factor for deep sky imaging rather than the telescope aperture...."
Here is what the Robust PCA decomposition provides:
Again, it looks like the specular reflection of the scene is picked up by the sparse component and the low rank component could be thought as some sort of averaging (but better) of the moon shot: i.e. comparable to a lucky imaging output but with fewer shots. The noisy component does indeed include the shot noise of the camera but also all the turbulence in the atmosphere. This is not noise, this information could be used as an aid for blind deconvolution and ultimately for other types of imaging such as Imaging with Nature ( see Imaging With Nature: Some thoughts on deblurring., Imaging With Nature, additional thoughts on deblurring, Old and New algorithm for Blind Deconvolution, Strike Number 5 in the "Technologies That Do Not Exist" Section )
Again, we did not need to set up a webcam and a telescope in our backyard to perform this experiment. And yes, the title is inspired from this scene in Batman, the movie..
Liked this entry ? subscribe to Nuit Blanche's feed, there's more where that came from. You can also subscribe to Nuit Blanche by Email, explore the Big Picture in Compressive Sensing or the Matrix Factorization Jungle and join the conversations on compressive sensing, advanced matrix factorization and calibration issues on Linkedin.