[ptx] Super-resolution images from many registered digicam images

[JD Smith]

		On Fri, 2004-09-17 at 03:03, Sebastian Nowozin wrote:
> Hello,
> 
> 
> I am dabbling with the idea to implement a resolution increase/fusion
> algorithm to combine two or more digicam images to one higher resolution
> image.

In the astronomical world, we do this all the time.  Many of the
spectacular images created on the Hubble Space Telescope are the result
of a form of super-resolution mosaicking embodied in an algorithm called
"Drizzle": search on this for more.  There are limits though.  The basic
upshot is that, if the "point spread function" (PSF), or in photography
terms, "circle of confusion" (COF), is well sampled by the detector
pixels, i.e. if the circle is 2 or more pixels in diameter, essentially
you'll not be able to recover any enhanced resolution.  If, however, as
in the case of many Hubble instruments, you are "under-sampled", with
less than 1-1.5 pixels per PSF/COF, you can recover some detail by
random sub-pixel dithering, and using drizzle to create a mosaic --
you're essentially fully sampling the PSF/COF in a series of
slightly-shifted exposures.

One simpler option to accomplish the same thing that is sometimes used
is to dither in a regular array of exactly 1/4 offset 1/2 pixel steps. 
Then you can easily recover 4 times the resolution (at max) by simply
interleaving the images on a 4x finer-sampled grid: nothing fancy
required.  

The first step of checking if this is practical would be to see if your
optics deliver a COF which is small compared to the pixel scale in the
detector.  If not (which will likely be the case for many systems),
there really isn't much room for any gains.

JD