Dave (and others)-
What do you think of taking the Fourier power spectrum of the images and
using as a definition of the noise those spatial frequecies at which the
power becomes constant then using that cut-off as a measure of when the
signal no longer exceeds noise. Then calling the spatial frequency where
signal is slightly greater than noise the resolution of the microscope in
that dimension.
________________________________________________________________________________
Paul Goodwin
Image Analysis Lab
FHCRC, Seattle, WA
On Wed, 1 Feb 1995, David W. Piston wrote:
> In response to Paul Goodwin's question about resolution:
>
> Any definition of resolution (Full-width Half-Maximum, Rayleigh's Criterion
> 1/e diameter, etc) is purely arbitrary. I see no reason that any would
> be, in general, poorly suited for electronic imaging. Certain definitions
> sometimes do fit a particular problem, though, and for calculations of
> background rejection and signal-to-noise in confocal microscopy, we used
> the 1/e^2 contour to define a focal volume. This volume corresponds to
> 90% of the light generated by a point source, so was convenient for our
> purposes (se articles by Sandison et al. in Applied Optics, vols. 35 & 37,
> and in the 3-D Confocal Microscopy book edited by Stevens, Mills & Trogadis).
> For your comparison between deconvolution and confocal, 1/2 widths seem
> as good as any other definition, but what you really need to compare is
> not resolution but information content (signal-to-noise).
>
> Unfortunately, this is not a trivial problem. . .
>
> Dave Piston
> Vanderbilt University
>
|