🚀 go-pugleaf

In article <1992Oct30.183655.14435@athena.mit.edu> ninassup@athena.mit.edu (Nikos I Nassuphis) writes:
>I am working on the controls of a device that
>includes a precision xy stage and a microscope
>connected to a camera and a framegrabber.
>
>I need to keep the camera focused onto a
>region of the surface as the stage moves it
>under the microscope.
>
>I therefore need some type of autofocus.
>I am contemplating a control loop that 
>will maximize the high frequency content of
>the area i want to focus on.
>
>Does anyone have any similar experience?
>Are there any relevant publications to look
>for?

You're on the right track.  I believe that some of the commercial
autofocus systems use a total energy maximization scheme, which is
basically the same as maximizing high-frequency energy in this case.
The nice thing about total energy maximization is that it's simpler to
evaluate than a frequency-domain expression.  Just maximize \sum
f^2(i,j) over focal distance.  This has a distinct maximum at the
proper focusing.  Note that \sum f(i,j) just represents the total
photon count and should not change with defocusing.  Thus, maximizing
\sum f^2(i,j) is equivalent to maximizing \frac{\sum f^2(i,j)}{[\sum
f(i,j)]^2}, which is the ratio of the energy in all frequencies over the
DC energy.  The only related reference I have is the following:

@article{muller74,
    TITLE = "Real-time correction of atmospherically degraded telescope
    images through image sharpening",
    AUTHOR = "R. A.  Muller and A. Buffington", 
	JOURNAL = "Journal of the Optical Society of America", 
	YEAR = "1974", 
	VOLUME = "64",
    NUMBER = "9", 
	PAGES = "1200-1210", 
	MONTH = "September"}

Hope that helps.



--
Stan Reeves
Auburn University, Department of Electrical Engineering, Auburn, AL  36849
INTERNET: sjreeves@eng.auburn.edu

In article <1992Oct30.183655.14435@athena.mit.edu> ninassup@athena.mit.edu  
(Nikos I Nassuphis) writes:
> I need to keep the camera focused onto a
> region of the surface as the stage moves it
> under the microscope.
> 
> I therefore need some type of autofocus.
> I am contemplating a control loop that 
> will maximize the high frequency content of
> the area i want to focus on.

You could try maximizing contrast. Michelson contrast is defined as 
(max - min)/(max + min) over some area. This will be computationally 
faster and easier to program than either maximizing high-frequency 
content or maximizing total energy of non-zero components. Whether or
not it will work will depend on what your image looks like.

Charles Neveu
Neurology Unit
Program in Physiological Optics
UCBerkeley