Autofocusing and image grey level mean

[Weller Andrew Francis]

Dear all,

Apologise for the long-winded introduction, but I'll get to my point...

Based on the conclusions drawn from Sun et al. 2004 (Autofocusing in
Computer Microscopy: Selecting the Optimal Focus Algorithm, MICROSCOPY
RESEARCH AND TECHNIQUE, 65, 139–149) I would like to run the 'Normalized
Variance' algorithm for autofocusing in microscopy.

My interpretation of the algorithm is as follows (please tell me if I am
wrong?!?):

a = 1 / ([image height] x [image width] x [image grey level mean
intensity])

b = ([pixel grey level] - [image grey level mean intensity])^2

Normalized Variance = a x b

(Where the best focus is defined as having the maximum value; this
decreases as defocus increases.)

My questions therefore are as follows:

By clicking Analyze > Measure I am given a 'Mean' value. I presume this
is the image's grey level mean intensity if no ROI is selected? If it is
not, how do I find and store this value so I use it to determine the
'Normalized Variance'?

Does anybody have any suggestions regarding a different/better
autofocusing approach?

Thanks, Andy

[Gabriel Landini]

On Wednesday 14 June 2006 16:03, Andy Weller wrote:
> My interpretation of the algorithm is as follows (please tell me if I am
> wrong?!?):
>
> a = 1 / ([image height] x [image width] x [image grey level mean
> intensity])
>
> b = ([pixel grey level] - [image grey level mean intensity])^2
>
> Normalized Variance = a x b

Not sure you mean what you wrote. b is calculated over a HxW window (or ROI),
so the variance is extracted from that window.
You have to accumulate the squared differences of each pixel in the window
with the mean intensity of the window. The sum of all that is b.
The focus is found when the variance is the largest, but this will not work
very well with noisy images.

I hope it helps,

G.