Posted by
Straatman, Kees (Dr.) on
URL: http://confocal-microscopy-list.275.s1.nabble.com/averaging-vs-accumulation-for-noise-reduction-is-there-a-difference-tp6483751p6487752.html
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy*****
Dear Jim and Guy,
Thanks for the answers. Should have thought before I wrote.
Kees
-----Original Message-----
From: Confocal Microscopy List [mailto:
[hidden email]] On Behalf Of James Pawley
Sent: 17 June 2011 15:18
To:
[hidden email]
Subject: Re: averaging vs. accumulation for noise reduction - is there a difference?
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy*****
Dear Kees,
In the confocal (or 2photon) microscope, their is very little
read-noise (noise signal generated by and within the photodetector,
in this case usually a PMT.). The main source of noise is Poisson
Noise which is a statistical noise, namely the fact that the
uncertainty of any measurement of quantum events is the square root
of the number of these events actually detected.
Suppose that you have a "constant signal" which, after being measured
many times, would average out to say, 100 photons/pixel. Any
particular measurement of this signal will be in the range of 90 to
110 photons about 67% of the time while 37% of the measurements are
even farther from the "real" intensity. We say that such a signal
has 10% Poisson noise.
On the other hand, confocal signals from the stained areas of the
specimen tend to be much less than 100 photons/pixel: often closer to
9 photon/pixel, giving +/-30% noise. Dark areas are likely to be 1's
and zeroes.
Multiplicative noise in the PMT (also related to Poisson statistics)
means that each photoelectron is not amplified by the same amount,
smearing out the intensity actually recorded. This means that even
though the recorded image of a weakly fluorescent object may seem to
have values everywhere from 0 to 256, you may in fact be only
recording a "true" signal of 1s and zeros with a few 2's thrown in by
the statistics. i.e., Unless you do photon counting ,you have little
idea of the actual number of photons represented by the number
recorded in the memory.
People commonly say that the signal gets "noisy" if they turn up the
PMT too far. While it is possible that their PMT is breaking down in
some way when used with higher voltages on the dynodes, the more
likely explanation is that they turn up the PMT because the signal is
so weak and weak signals have very poor Poisson statistics.
Counting longer or Kalman averaging allows more photons/pixel to be
counted. Although counting more events increases the absolute
uncertainty, (the sqrt of a larger number is bigger than the sqrt of
a smaller number), the ratio of the signal to this noise improves
with the sqrt of the total signal detected.
To bring another variable into the discussion, deconvolving confocal
or 2-photon data effectively averages the signal over all the voxels
needed to define the PSF. If you assume 2 voxels are needed for each
"resel" (the resolution element, from the center of the Airy Disk to
the first dark ring on one side), then you need at least 4x4x4 = 64
voxels to define the PSF, and deconvolving such a signal in 3D will
effectively average the signal of a point object over these 64
nearest neighbors, drastically improving the S/N (In this case the
S/N would improve 8x, the same as averaging for 64 scans. The
contrast of the image would also appear to decrease but this is an
artefact of averaging and can be increased using a look-up table.)
Hope that helps.
Jim Pawley
at the 2011 UBC Course.
>
>Very interesting discussion.
>
>I was just wondering. We talk about S/N. If you acquire a single
>frame for longer I would expect you also accumulate more noise in
>the image, if you average your signal in each image your sample
>should be more or less the same while the noise (random) is
>averaged. So would you not expect a better S/N with averaging when
>all else is equal?
>
>Kees
>
>Senior Experimental Officer
>Centre for Core Biotechnology Services
>University of Leicester, UK
>
>
http://www.le.ac.uk/biochem/microscopy/home.html>
>
>-----Original Message-----
>From: Confocal Microscopy List
>[mailto:
[hidden email]] On Behalf Of Stanislav
>Vitha
>Sent: 16 June 2011 17:26
>To:
[hidden email]
>Subject: averaging vs. accumulation for noise reduction - is there a
>difference?
>
>*****
>To join, leave or search the confocal microscopy listserv, go to:
>
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy>*****
>
>Hallo,
>this is a very basic question, but I cannot figure this out from what I have
>been reading, so a simple explanation for a non-physicist would be much
>appreciated:
>
>Is there a real difference in the improvement of the signal to noise ratio
>between frame averaging (or accumulation) and longer dwell times (slower
>scan) for a point-scanning confocal witrh a PMT detector?
>
>For instance, using single point scanning confocal, 12-bit acquisition.
>
>a) averaging (or accumulating) 5 frames, 4 microseconds per pixel
>b) acquiring a single frame, 20 microseconds per pixel
>
>Assumptions:
>no saturation of the detector;
>stable environmental conditions, no focus drift, etc
>
>Would it matter (for the dfference between the two scenarios) if it was analog
>detection or photon counting detection?
>
>I will run this little test later, but I am curious what you think.
>
>I thought that at least for the photon counting mode, the two important
>factors would be the dark counts and the total number of counts detected, so
>whether it is acquired in one scan or in 5 scans, it should be the same. My
>camera expert here insists that the averaging scheme will give better noise
>suppression.
>
>Thanks!
>
>
>Stan Vitha
>
>Microscopy and Imaging Center
>Texas A&M University
--
****************************************
Prof. James B. Pawley, Phone: 604-822-7801
3D Microscopy of Living Cells: Summer Course CELL: 778-919-3176
Info at:
http://www.3dcourse.ubc.ca"If it isn't diffraction, it is statistics":Microscopist's complaint, Anon.