Confocal Microscopy List

median filtering confocal microscope data at the instrument

Classic

List

Threaded

21 messages Options

George McNamara

median filtering confocal microscope data at the instrument

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Dear confocal listserv,

While training a new user on our Zeiss LSM710 confocal microscope, I
tried explaining the differences between Kalman filtering, arithmetic
averaging, and my suggestion (see the archive for posts) of using the
median for each pixel (assuming no photobleaching). The main intent
being that under situations with most of the noise being from the PMT,
that the probability of the PMT having a high value in 3 of 5 time
points for a given pixel would be low (unless the PMT high voltage was
cranked really high). It occurred to me that the Zeiss ZEN software
includes median filtering. I acquired a 5 timepoint series of a single
plane, went to the ZEN (2010B SP1) selected median filter, X=1, Y=1,
Z=5, selected the image (since Zeiss is too stupid to figure out that
the current image is probably what the user wants to filter - especially
if it is the only open image), clicked apply, then after the result
image series was created, changed to plane 3. Result was: nice (planes
1, 2, 4 and 5 were not - this also shows that Zeiss does not know Z from
time). This particular user's specimen autofluorescence and/or
non-specific antibodies was a more significant factor in their
experiment than the PMT gain, but I suggest median filtering will be
useful in some situations. I encourage confocal vendors (EMCCd data
might also benefit from something like it, median over an odd number of
consecutive frames) make this an option. Certainly more sensible in 2013
than Kalman filtering.

Enjoy,

George

Jeremy Adler-4

Re: median filtering confocal microscope data at the instrument

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

I dislike median filtering since by taking the median (only the
central value) value most the data is thrown away. Normally Poisson
noise is the problem and the sum of the timeseries will have a lower
uncertainty than the median.

In the instance George cites, of occasional substantial noise from the
PMT, a better option is decide whether the distribution of five values
is likely to include aberrant high value(s) and if so take the
arithmetic of the remaining values. The problem is to decide if the
distribution of the 5 values is consistent with Poisson noise, which
is tricky. However over the whole image there is a population of
pixels and for each mean the likely population distribution could be
generated - aberrant values could then be detected with higher
probability.

Quoting George McNamara <[hidden email]>:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Dear confocal listserv,
>
> While training a new user on our Zeiss LSM710 confocal microscope, I
> tried explaining the differences between Kalman filtering,
> arithmetic averaging, and my suggestion (see the archive for posts)
> of using the median for each pixel (assuming no photobleaching). The
> main intent being that under situations with most of the noise being
> from the PMT, that the probability of the PMT having a high value in
> 3 of 5 time points for a given pixel would be low (unless the PMT
> high voltage was cranked really high). It occurred to me that the
> Zeiss ZEN software includes median filtering. I acquired a 5
> timepoint series of a single plane, went to the ZEN (2010B SP1)
> selected median filter, X=1, Y=1, Z=5, selected the image (since
> Zeiss is too stupid to figure out that the current image is probably
> what the user wants to filter - especially if it is the only open
> image), clicked apply, then after the result image series was
> created, changed to plane 3. Result was: nice (planes 1, 2, 4 and 5
> were not - this also shows that Zeiss does not know Z from time).
> This particular user's specimen autofluorescence and/or non-specific
> antibodies was a more significant factor in their experiment than
> the PMT gain, but I suggest median filtering will be useful in some
> situations. I encourage confocal vendors (EMCCd data might also
> benefit from something like it, median over an odd number of
> consecutive frames) make this an option. Certainly more sensible in
> 2013 than Kalman filtering.
>
> Enjoy,
>
> George
>

Jeremy Adler
IGP
Rudbeckslaboratoriet
Daghammersköljdsväg 20
751 85 Uppsala
Sweden

0046 (0)18 471 4607

Guy Cox-2

Re: median filtering confocal microscope data at the instrument

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

I agree with the general tenor of this post. But I disagree with the statement that a median filter 'throws away' most of the data. All the pixels within the kernel contribute to the final result - and then the next set contributes to the next pixel, and so on. It is in fact quite a good way of doing just what Jeremy wants - getting rid of outliers. This is not to deny that more complex algorithms might not do even better. But it does not throw away data.

Guy

-----Original Message-----
From: Confocal Microscopy List [mailto:[hidden email]] On Behalf Of Jeremy Adler
Sent: Monday, 25 February 2013 4:09 AM
To: [hidden email]
Subject: Re: median filtering confocal microscope data at the instrument

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

I dislike median filtering since by taking the median (only the central value) value most the data is thrown away. Normally Poisson noise is the problem and the sum of the timeseries will have a lower uncertainty than the median.

In the instance George cites, of occasional substantial noise from the PMT, a better option is decide whether the distribution of five values is likely to include aberrant high value(s) and if so take the arithmetic of the remaining values. The problem is to decide if the distribution of the 5 values is consistent with Poisson noise, which is tricky. However over the whole image there is a population of pixels and for each mean the likely population distribution could be generated - aberrant values could then be detected with higher probability.

Quoting George McNamara <[hidden email]>:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Dear confocal listserv,
>
> While training a new user on our Zeiss LSM710 confocal microscope, I
> tried explaining the differences between Kalman filtering, arithmetic
> averaging, and my suggestion (see the archive for posts) of using the
> median for each pixel (assuming no photobleaching). The main intent
> being that under situations with most of the noise being from the PMT,
> that the probability of the PMT having a high value in
> 3 of 5 time points for a given pixel would be low (unless the PMT high
> voltage was cranked really high). It occurred to me that the Zeiss ZEN
> software includes median filtering. I acquired a 5 timepoint series of
> a single plane, went to the ZEN (2010B SP1) selected median filter,
> X=1, Y=1, Z=5, selected the image (since Zeiss is too stupid to figure
> out that the current image is probably what the user wants to filter -
> especially if it is the only open image), clicked apply, then after
> the result image series was created, changed to plane 3. Result was:
> nice (planes 1, 2, 4 and 5 were not - this also shows that Zeiss does
> not know Z from time).
> This particular user's specimen autofluorescence and/or non-specific
> antibodies was a more significant factor in their experiment than the
> PMT gain, but I suggest median filtering will be useful in some
> situations. I encourage confocal vendors (EMCCd data might also
> benefit from something like it, median over an odd number of
> consecutive frames) make this an option. Certainly more sensible in
> 2013 than Kalman filtering.
>
> Enjoy,
>
> George
>

Jeremy Adler
IGP
Rudbeckslaboratoriet
Daghammersköljdsväg 20
751 85 Uppsala
Sweden

0046 (0)18 471 4607

Johannes Schindelin

Re: median filtering confocal microscope data at the instrument

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Dear Guy,

On Mon, 25 Feb 2013, Guy Cox wrote:

> I agree with the general tenor of this post. But I disagree with the
> statement that a median filter 'throws away' most of the data.

While I am not an enemy of the median filter, it has to be stated that it
throws away most of the data, even to the point of introducing artifacts.

> All the pixels within the kernel contribute to the final result - and
> then the next set contributes to the next pixel, and so on.

The problem is not so much that the pixels contribute or not, the problem
with the Median filter in particular is that it operates on an ordinal
scale, not an interval scale.

In other words, the neighboring pixels' contribution is *qualitative*, not
*quantitative*.

> It is in fact quite a good way of doing just what Jeremy wants - getting
> rid of outliers.

The Median filter is indeed a quite good way to filter out outliers of
data in a quick and robust manner.

The problem is that what you *should* want to do after acquiring images is
to analyze them *quantitatively*. And since the Median filter is a
non-linear filter, it is inappropriate to apply it before any quantitative
analysis.

> This is not to deny that more complex algorithms might not do even
> better. But it does not throw away data.

Sorry to re-iterate: it does throw away data. Just as almost every image
processing filter, the information (as can be measured by the entropy) is
reduced. The hope, of course, is that information was lost that is of no
interest (in this particular case, values from outliers), but there is
almost always also information lost that one wanted (in this particular
case, the linear relation between the original and the processed data).

Having said that, the Median filter might still be the most appropriate
filter for Jeremy's application -- unfortunately.

Ciao,
Johannes

George McNamara

Re: median filtering confocal microscope data at the instrument

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi Johannes,

So if a single pixel scanned five times has values of 40, 38, 42, 3800,
37, (not necessarily acquired in that order), you would prefer the
arithmetic mean 791.4 (not that any of the vendors can give you the 0.4,
and for anyone who is a fan of Kalman, the Kalman value would depend on
which order the five were acquired in, rather than 40 (which is the
"digital offset" I usually use on the Zeiss LSM710 I manage when
operating in 12-bit mode).

Pesonally, I would like to see the point scanning confocal microscope
(and EMCCD software) vendors implement median and even more PMT and
similar noisy data appropriate methods to provide the best possible data
to my users and I (and as of April: my colleagues and I at MDACC, Houston).

Sincerely,

George

On 2/25/2013 2:41 PM, Johannes Schindelin wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Dear Guy,
>
> On Mon, 25 Feb 2013, Guy Cox wrote:
>
>
>> I agree with the general tenor of this post. But I disagree with the
>> statement that a median filter 'throws away' most of the data.
>>
> While I am not an enemy of the median filter, it has to be stated that it
> throws away most of the data, even to the point of introducing artifacts.
>
>
>> All the pixels within the kernel contribute to the final result - and
>> then the next set contributes to the next pixel, and so on.
>>
> The problem is not so much that the pixels contribute or not, the problem
> with the Median filter in particular is that it operates on an ordinal
> scale, not an interval scale.
>
> In other words, the neighboring pixels' contribution is *qualitative*, not
> *quantitative*.
>
>
>> It is in fact quite a good way of doing just what Jeremy wants - getting
>> rid of outliers.
>>
> The Median filter is indeed a quite good way to filter out outliers of
> data in a quick and robust manner.
>
> The problem is that what you *should* want to do after acquiring images is
> to analyze them *quantitatively*. And since the Median filter is a
> non-linear filter, it is inappropriate to apply it before any quantitative
> analysis.
>
>
>> This is not to deny that more complex algorithms might not do even
>> better. But it does not throw away data.
>>
> Sorry to re-iterate: it does throw away data. Just as almost every image
> processing filter, the information (as can be measured by the entropy) is
> reduced. The hope, of course, is that information was lost that is of no
> interest (in this particular case, values from outliers), but there is
> almost always also information lost that one wanted (in this particular
> case, the linear relation between the original and the processed data).
>
> Having said that, the Median filter might still be the most appropriate
> filter for Jeremy's application -- unfortunately.
>
> Ciao,
> Johannes
>
>

Johannes Schindelin

Re: median filtering confocal microscope data at the instrument

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi George,

On Mon, 25 Feb 2013, George McNamara wrote:

> So if a single pixel scanned five times has values of 40, 38, 42, 3800,
> 37, (not necessarily acquired in that order), you would prefer the
> arithmetic mean 791.4 (not that any of the vendors can give you the 0.4,
> and for anyone who is a fan of Kalman, the Kalman value would depend on
> which order the five were acquired in, rather than 40 (which is the
> "digital offset" I usually use on the Zeiss LSM710 I manage when
> operating in 12-bit mode).

Please do not pretend that I said that. When I talked about "linear"
filters, I did not imply a simple averaging.

> Pesonally, I would like to see the point scanning confocal microscope
> (and EMCCD software) vendors implement median and even more PMT and
> similar noisy data appropriate methods to provide the best possible data
> to my users and I (and as of April: my colleagues and I at MDACC,
> Houston).

What data would be best recorded depends highly on the application. In the
general case, recording the values 40, 38, 42, 3800 and 37 in your above
example would be better than recording just "40". But recording more than
one value per pixel is often not practical.

To reiterate: The Median filter *can* be the optimal filter. You should
just not go around and tell everybody that it *is* the optimal filter,
because it certainly is not. And in particular when you want to quantify
your data after acquisition, it is inappropriate to use the Median filter
(remember: a filter should not be applied just because the processed image
"looks good", but it should only be applied if it helps the analysis).

That is all I said. (I certainly did not claim that you should always take
a simple arithmetic mean. I am not that stupid.)

Ciao,
Johannes

James Pawley

Re: median filtering confocal microscope data at the instrument

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi all,

It seems that we are discussing the best ways of eliminating the
effects of what are sometimes called "single-pixel" noise events.
Although it is fair to ask "What other kind of noise is there?", the
term is often used to refer to pixels with recorded intensity values
that are "unreasonably large" and seem to have nothing to do with the
presence of dye molecules at a certain location in the specimen. Such
values can come from a number of possible sources: cosmic rays pass
through the photocathode every few seconds; alpha particles from
radioactive elements in the PMT somewhat more often. If the PMT is
used at high gain, as is often the case when looking at living
specimens and signal levels must be kept low, single-photoelectron
dark counts may produce fast pulses from the PMT or EM-CCD that
approach the size of those representing signal in a "stained" pixel.
As these signals seem obvious artifacts when viewed by eye, it would
be convenient if they could be removed automatically.

By definition, filters take things out. That is both their aim and
their curse. One can argue for hours on whether or not the resulting
data is better or not. In addition, filters are very fast. However,
as computers get ever faster and cheaper one would expect that this
advantage would become less important.

In contrast to filters, deconvolution puts things in. Traditionally,
it reimposes the limits known to have been placed on the data by the
optics used to obtain it. Single-pixel events are "impossible"
because, assuming that Nyquist has been satisfied, the smallest
"real" feature in the data should be at least 4 pixels wide (or 12-16
pixels in area, 50-100 voxels in volume), not one pixel.

Because the spatial frequency of a noise pulse singularity is at
least 4x higher than that of the highest spatial frequency that the
optical system is capable of having transmitted, the offending value
can be tagged and then either replaced or averaged down. Indeed, some
EM-CCDs can now be set to detect and remove single-pixel noise based
on this recognition.

More generally however, the most reliable and robust method of
obtaining the most accurate pixel-intensity information from a series
of sequentially-obtained data sets is to deconvolve them in time as
well as space. This just means that we put into the process not just
the PSF (which sets the limits on possible spatial frequencies) but
also our knowledge that real changes in specimen brightness can only
occur so fast and not any faster.

George has postulated a series of intensity measurements from a
single pixel. Depending on the time delay between the measurements in
this series, we may (or may not) be justified in assuming that no
real change in the specimen could justify a sudden, 100x intensity
change that is only one scan-time in duration. Again, this allows us
to tag outliers and then dispose of them either by replacement or
averaging.

So much for what can now be conveniently done using computers.

Let us not forget that every effort should also be taken to reduce
the number of single-pixel anomalies present in the raw data to begin
with. With the PMT, this means keeping the photocathode cool and
small and monitoring its no-real-signal output over time (i.e., dark
count rate). Store a reference image of a single scan with all the
lasers and room lights turned off, and look for changes in its
general appearance as the weeks pass. More quantitative measures are
also wise. (And while you are at it, compare this zero-light result
with one obtained when the level of room illumination present is
similar to that which you use when actually collecting data. Stray
light is often a more serious problem than we we expect.)

When employing an EM-CCD, a similar no-signal image can be used to
assess changes in dark-count and coupling-induced charge over time.
These may slowly drift up over time (months) and are always very
sensitive to chip temperature.

I am less familiar with the anomalous, single-pixel behavior of sCMOS
cameras, but I would guess that, with the exception of hot pixels,
they are less common simply because charge amplification is not
involved and events associated with the emergence of a single, errant
photoelectron cannot be seen above the general read-noise level. As
hot-pixels tend to reoccur at the same exact location in the image,
they can be automatically identified and averaged out using data from
their 4 or 8 nearest neighbours.

In all these cases the most efficient way of removing single-pixel
anomalies from your final data is to take all the precautions needed
to prevent them from occurring.

Sorry for being so long-winded.

Jim Pawley

>*****
>To join, leave or search the confocal microscopy listserv, go to:
>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>*****
>
>Hi George,
>
>On Mon, 25 Feb 2013, George McNamara wrote:
>
>> So if a single pixel scanned five times has values of 40, 38, 42, 3800,
>> 37, (not necessarily acquired in that order), you would prefer the
>> arithmetic mean 791.4 (not that any of the vendors can give you the 0.4,
>> and for anyone who is a fan of Kalman, the Kalman value would depend on
>> which order the five were acquired in, rather than 40 (which is the
>> "digital offset" I usually use on the Zeiss LSM710 I manage when
>> operating in 12-bit mode).
>
>Please do not pretend that I said that. When I talked about "linear"
>filters, I did not imply a simple averaging.
>
>> Pesonally, I would like to see the point scanning confocal microscope
>> (and EMCCD software) vendors implement median and even more PMT and
>> similar noisy data appropriate methods to provide the best possible data
>> to my users and I (and as of April: my colleagues and I at MDACC,
>> Houston).
>
>What data would be best recorded depends highly on the application. In the
>general case, recording the values 40, 38, 42, 3800 and 37 in your above
>example would be better than recording just "40". But recording more than
>one value per pixel is often not practical.
>
>To reiterate: The Median filter *can* be the optimal filter. You should
>just not go around and tell everybody that it *is* the optimal filter,
>because it certainly is not. And in particular when you want to quantify
>your data after acquisition, it is inappropriate to use the Median filter
>(remember: a filter should not be applied just because the processed image
>"looks good", but it should only be applied if it helps the analysis).
>
>That is all I said. (I certainly did not claim that you should always take
>a simple arithmetic mean. I am not that stupid.)
>
>Ciao,
>Johannes

--
James and Christine Pawley, 5446 Burley Place (PO Box 2348), Sechelt,
BC, Canada, V0N3A0,
Phone 604-885-0840, email <[hidden email]>
NEW! NEW! AND DIFFERENT Cell (when I remember to turn it on!) 1-604-989-6146

George McNamara

Re: median filtering confocal microscope data at the instrument

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi Jim et al,

I like your point of being able to operate in time as well as in XY and
Z. For that matter, if FLIM detectors become widespread, they operate at
relatively high gain, so should have a very different noise signature
than fluorescence lifetimes. This could - and should - be cleaned up!
dead time and other issues could also be modelled and dealt with).

Except for some EMCCD "hot pixel' noise removal - which sounds to me
like a marketing feature, none of these denoising concepts have been
implemented by any of the big four laser scanning confocal microscope
vendors. I am pretty confident that there are over 1500 Leica SP5's in
the field (and a few SP8's, and some unfortunate souls still nursing
SP1's and SP2's), and similar number (or more) of Zeiss LSM
510+700+710+780's. My guess is Nikon + Olympus together add up to around
1500, maybe more. So, market size of around 5,000 systems. I know Leica
does not have online median (or any other) pixel cleanup in the LAS AF
software I have on my SP5's (and the "batch deconvolution" is so awful
even Leica's applications scientists have never been able to show me it
working). I discovered that Zeiss ZEN could be tricked into doing a
median filter on time series data (post acquisition, and tedious number
of steps to do it). I doubt Nikon or Olympus software has it.

Upshot: none of the major vendors has made pixel cleanup a
consideration. Come to think of it, I don't know of any of the minor
vendors having clean up either.

I do want to remind readers that the PMT noise I am referring to is most
relevant when specimen autofluorescence is very low - or when the lasers
are blocked so you can evaluate the detection side without pesky photons
messing up your head.

Enjoy,

George

On 2/26/2013 11:45 AM, James Pawley wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Hi all,
>
> It seems that we are discussing the best ways of eliminating the
> effects of what are sometimes called "single-pixel" noise events.
> Although it is fair to ask "What other kind of noise is there?", the
> term is often used to refer to pixels with recorded intensity values
> that are "unreasonably large" and seem to have nothing to do with the
> presence of dye molecules at a certain location in the specimen. Such
> values can come from a number of possible sources: cosmic rays pass
> through the photocathode every few seconds; alpha particles from
> radioactive elements in the PMT somewhat more often. If the PMT is
> used at high gain, as is often the case when looking at living
> specimens and signal levels must be kept low, single-photoelectron
> dark counts may produce fast pulses from the PMT or EM-CCD that
> approach the size of those representing signal in a "stained" pixel.
> As these signals seem obvious artifacts when viewed by eye, it would
> be convenient if they could be removed automatically.
>
> By definition, filters take things out. That is both their aim and
> their curse. One can argue for hours on whether or not the resulting
> data is better or not. In addition, filters are very fast. However, as
> computers get ever faster and cheaper one would expect that this
> advantage would become less important.
>
> In contrast to filters, deconvolution puts things in. Traditionally,
> it reimposes the limits known to have been placed on the data by the
> optics used to obtain it. Single-pixel events are "impossible"
> because, assuming that Nyquist has been satisfied, the smallest "real"
> feature in the data should be at least 4 pixels wide (or 12-16 pixels
> in area, 50-100 voxels in volume), not one pixel.
>
> Because the spatial frequency of a noise pulse singularity is at least
> 4x higher than that of the highest spatial frequency that the optical
> system is capable of having transmitted, the offending value can be
> tagged and then either replaced or averaged down. Indeed, some EM-CCDs
> can now be set to detect and remove single-pixel noise based on this
> recognition.
>
> More generally however, the most reliable and robust method of
> obtaining the most accurate pixel-intensity information from a series
> of sequentially-obtained data sets is to deconvolve them in time as
> well as space. This just means that we put into the process not just
> the PSF (which sets the limits on possible spatial frequencies) but
> also our knowledge that real changes in specimen brightness can only
> occur so fast and not any faster.
>
> George has postulated a series of intensity measurements from a single
> pixel. Depending on the time delay between the measurements in this
> series, we may (or may not) be justified in assuming that no real
> change in the specimen could justify a sudden, 100x intensity change
> that is only one scan-time in duration. Again, this allows us to tag
> outliers and then dispose of them either by replacement or averaging.
>
> So much for what can now be conveniently done using computers.
>
> Let us not forget that every effort should also be taken to reduce the
> number of single-pixel anomalies present in the raw data to begin
> with. With the PMT, this means keeping the photocathode cool and small
> and monitoring its no-real-signal output over time (i.e., dark count
> rate). Store a reference image of a single scan with all the lasers
> and room lights turned off, and look for changes in its general
> appearance as the weeks pass. More quantitative measures are also
> wise. (And while you are at it, compare this zero-light result with
> one obtained when the level of room illumination present is similar to
> that which you use when actually collecting data. Stray light is often
> a more serious problem than we we expect.)
>
> When employing an EM-CCD, a similar no-signal image can be used to
> assess changes in dark-count and coupling-induced charge over time.
> These may slowly drift up over time (months) and are always very
> sensitive to chip temperature.
>
> I am less familiar with the anomalous, single-pixel behavior of sCMOS
> cameras, but I would guess that, with the exception of hot pixels,
> they are less common simply because charge amplification is not
> involved and events associated with the emergence of a single, errant
> photoelectron cannot be seen above the general read-noise level. As
> hot-pixels tend to reoccur at the same exact location in the image,
> they can be automatically identified and averaged out using data from
> their 4 or 8 nearest neighbours.
>
> In all these cases the most efficient way of removing single-pixel
> anomalies from your final data is to take all the precautions needed
> to prevent them from occurring.
>
> Sorry for being so long-winded.
>
> Jim Pawley
>
>
>> *****
>> To join, leave or search the confocal microscopy listserv, go to:
>> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>> *****
>>
>> Hi George,
>>
>> On Mon, 25 Feb 2013, George McNamara wrote:
>>
>>> So if a single pixel scanned five times has values of 40, 38, 42,
>>> 3800,
>>> 37, (not necessarily acquired in that order), you would prefer the
>>> arithmetic mean 791.4 (not that any of the vendors can give you the
>>> 0.4,
>>> and for anyone who is a fan of Kalman, the Kalman value would
>>> depend on
>>> which order the five were acquired in, rather than 40 (which is the
>>> "digital offset" I usually use on the Zeiss LSM710 I manage when
>>> operating in 12-bit mode).
>>
>> Please do not pretend that I said that. When I talked about "linear"
>> filters, I did not imply a simple averaging.
>>
>>> Pesonally, I would like to see the point scanning confocal microscope
>>> (and EMCCD software) vendors implement median and even more PMT and
>>> similar noisy data appropriate methods to provide the best possible
>>> data
>>> to my users and I (and as of April: my colleagues and I at MDACC,
>>> Houston).
>>
>> What data would be best recorded depends highly on the application.
>> In the
>> general case, recording the values 40, 38, 42, 3800 and 37 in your above
>> example would be better than recording just "40". But recording more
>> than
>> one value per pixel is often not practical.
>>
>> To reiterate: The Median filter *can* be the optimal filter. You should
>> just not go around and tell everybody that it *is* the optimal filter,
>> because it certainly is not. And in particular when you want to quantify
>> your data after acquisition, it is inappropriate to use the Median
>> filter
>> (remember: a filter should not be applied just because the processed
>> image
>> "looks good", but it should only be applied if it helps the analysis).
>>
>> That is all I said. (I certainly did not claim that you should always
>> take
>> a simple arithmetic mean. I am not that stupid.)
>>
>> Ciao,
>> Johannes
>
>

mcammer

Re: median filtering confocal microscope data at the instrument

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

When it comes to software to deal with confocal datasets there is one that we rely on for filtering in time and space and doing quantification and much of image display in a highly flexible programmable environment: ImageJ.

_________________________________________
Michael Cammer, Assistant Research Scientist
Skirball Institute of Biomolecular Medicine
Lab: (212) 263-3208 Cell: (914) 309-3270

________________________________________
From: Confocal Microscopy List [[hidden email]] on behalf of George McNamara [[hidden email]]
Sent: Tuesday, February 26, 2013 9:06 PM
To: [hidden email]
Subject: Re: median filtering confocal microscope data at the instrument

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi Jim et al,

I like your point of being able to operate in time as well as in XY and
Z. For that matter, if FLIM detectors become widespread, they operate at
relatively high gain, so should have a very different noise signature
than fluorescence lifetimes. This could - and should - be cleaned up!
dead time and other issues could also be modelled and dealt with).

Except for some EMCCD "hot pixel' noise removal - which sounds to me
like a marketing feature, none of these denoising concepts have been
implemented by any of the big four laser scanning confocal microscope
vendors. I am pretty confident that there are over 1500 Leica SP5's in
the field (and a few SP8's, and some unfortunate souls still nursing
SP1's and SP2's), and similar number (or more) of Zeiss LSM
510+700+710+780's. My guess is Nikon + Olympus together add up to around
1500, maybe more. So, market size of around 5,000 systems. I know Leica
does not have online median (or any other) pixel cleanup in the LAS AF
software I have on my SP5's (and the "batch deconvolution" is so awful
even Leica's applications scientists have never been able to show me it
working). I discovered that Zeiss ZEN could be tricked into doing a
median filter on time series data (post acquisition, and tedious number
of steps to do it). I doubt Nikon or Olympus software has it.

Upshot: none of the major vendors has made pixel cleanup a
consideration. Come to think of it, I don't know of any of the minor
vendors having clean up either.

Xuelin

Re: median filtering confocal microscope data at the instrument

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Dear List

could someone tell me what's difference between opening pinhole and median
filter.
thanks

2013/2/27 Cammer, Michael <[hidden email]>ddi

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> When it comes to software to deal with confocal datasets there is one that
> we rely on for filtering in time and space and doing quantification and
> much of image display in a highly flexible programmable environment:
> ImageJ.
>
> _________________________________________
> Michael Cammer, Assistant Research Scientist
> Skirball Institute of Biomolecular Medicine
> Lab: (212) 263-3208 Cell: (914) 309-3270
>
> ________________________________________
> From: Confocal Microscopy List [[hidden email]] on
> behalf of George McNamara [[hidden email]]
> Sent: Tuesday, February 26, 2013 9:06 PM
> To: [hidden email]
> Subject: Re: median filtering confocal microscope data at the instrument
>
> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Hi Jim et al,
>
> I like your point of being able to operate in time as well as in XY and
> Z. For that matter, if FLIM detectors become widespread, they operate at
> relatively high gain, so should have a very different noise signature
> than fluorescence lifetimes. This could - and should - be cleaned up!
> dead time and other issues could also be modelled and dealt with).
>
> Except for some EMCCD "hot pixel' noise removal - which sounds to me
> like a marketing feature, none of these denoising concepts have been
> implemented by any of the big four laser scanning confocal microscope
> vendors. I am pretty confident that there are over 1500 Leica SP5's in
> the field (and a few SP8's, and some unfortunate souls still nursing
> SP1's and SP2's), and similar number (or more) of Zeiss LSM
> 510+700+710+780's. My guess is Nikon + Olympus together add up to around
> 1500, maybe more. So, market size of around 5,000 systems. I know Leica
> does not have online median (or any other) pixel cleanup in the LAS AF
> software I have on my SP5's (and the "batch deconvolution" is so awful
> even Leica's applications scientists have never been able to show me it
> working). I discovered that Zeiss ZEN could be tricked into doing a
> median filter on time series data (post acquisition, and tedious number
> of steps to do it). I doubt Nikon or Olympus software has it.
>
> Upshot: none of the major vendors has made pixel cleanup a
> consideration. Come to think of it, I don't know of any of the minor
> vendors having clean up either.
>

--
Best, 赵学林

Chris Tully

Re: median filtering confocal microscope data at the instrument

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Xuelin,

A pinhole restricts the light coming from the sample, allowing only
the zero order diffraction peak to reach the sensor.

A median filter is a post acquisition process that looks at some
neighborhood of pixels and assigns the median value of that
neighborhood to the center pixel.

Chris Tully, M.S., Image Analysis Expert
t 240.475.9753 f 419.831.0527 | [hidden email]

Sent from my iPhone please excuse typos.

On Feb 27, 2013, at 4:41 AM, xuelin zhao <[hidden email]> wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Dear List
>
> could someone tell me what's difference between opening pinhole and median
> filter.
> thanks
>
> 2013/2/27 Cammer, Michael <[hidden email]>ddi
>
>> *****
>> To join, leave or search the confocal microscopy listserv, go to:
>> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>> *****
>>
>> When it comes to software to deal with confocal datasets there is one that
>> we rely on for filtering in time and space and doing quantification and
>> much of image display in a highly flexible programmable environment:
>> ImageJ.
>>
>> _________________________________________
>> Michael Cammer, Assistant Research Scientist
>> Skirball Institute of Biomolecular Medicine
>> Lab: (212) 263-3208 Cell: (914) 309-3270
>>
>> ________________________________________
>> From: Confocal Microscopy List [[hidden email]] on
>> behalf of George McNamara [[hidden email]]
>> Sent: Tuesday, February 26, 2013 9:06 PM
>> To: [hidden email]
>> Subject: Re: median filtering confocal microscope data at the instrument
>>
>> *****
>> To join, leave or search the confocal microscopy listserv, go to:
>> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>> *****
>>
>> Hi Jim et al,
>>
>> I like your point of being able to operate in time as well as in XY and
>> Z. For that matter, if FLIM detectors become widespread, they operate at
>> relatively high gain, so should have a very different noise signature
>> than fluorescence lifetimes. This could - and should - be cleaned up!
>> dead time and other issues could also be modelled and dealt with).
>>
>> Except for some EMCCD "hot pixel' noise removal - which sounds to me
>> like a marketing feature, none of these denoising concepts have been
>> implemented by any of the big four laser scanning confocal microscope
>> vendors. I am pretty confident that there are over 1500 Leica SP5's in
>> the field (and a few SP8's, and some unfortunate souls still nursing
>> SP1's and SP2's), and similar number (or more) of Zeiss LSM
>> 510+700+710+780's. My guess is Nikon + Olympus together add up to around
>> 1500, maybe more. So, market size of around 5,000 systems. I know Leica
>> does not have online median (or any other) pixel cleanup in the LAS AF
>> software I have on my SP5's (and the "batch deconvolution" is so awful
>> even Leica's applications scientists have never been able to show me it
>> working). I discovered that Zeiss ZEN could be tricked into doing a
>> median filter on time series data (post acquisition, and tedious number
>> of steps to do it). I doubt Nikon or Olympus software has it.
>>
>> Upshot: none of the major vendors has made pixel cleanup a
>> consideration. Come to think of it, I don't know of any of the minor
>> vendors having clean up either.
>
>
>
> --
> Best, 赵学林

Guy Cox-2

Re: median filtering confocal microscope data at the instrument

In reply to this post by James Pawley

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Jim,

OK, we are probably going to come to blows over this. I just trust the buffer of the Pacific Ocean between us. The term 'filter' applied to digital operations is a bit unfortunate. An optical filter removes light according to its specification. A digital, so called, filter does nothing of the sort. It processes pixels according to the values of other pixels. Deconvolution does EXACTLY the same thing - just with a more sophisticated algorithm. Fundamentally there is no difference. I really wish the term 'filter' had never been used in the digital world.

Guy

-----Original Message-----
From: Confocal Microscopy List [mailto:[hidden email]] On Behalf Of James Pawley
Sent: Wednesday, 27 February 2013 3:45 AM
To: [hidden email]
Subject: Re: median filtering confocal microscope data at the instrument

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi all,

It seems that we are discussing the best ways of eliminating the
effects of what are sometimes called "single-pixel" noise events.
Although it is fair to ask "What other kind of noise is there?", the
term is often used to refer to pixels with recorded intensity values
that are "unreasonably large" and seem to have nothing to do with the
presence of dye molecules at a certain location in the specimen. Such
values can come from a number of possible sources: cosmic rays pass
through the photocathode every few seconds; alpha particles from
radioactive elements in the PMT somewhat more often. If the PMT is
used at high gain, as is often the case when looking at living
specimens and signal levels must be kept low, single-photoelectron
dark counts may produce fast pulses from the PMT or EM-CCD that
approach the size of those representing signal in a "stained" pixel.
As these signals seem obvious artifacts when viewed by eye, it would
be convenient if they could be removed automatically.

By definition, filters take things out. That is both their aim and
their curse. One can argue for hours on whether or not the resulting
data is better or not. In addition, filters are very fast. However,
as computers get ever faster and cheaper one would expect that this
advantage would become less important.

In contrast to filters, deconvolution puts things in. Traditionally,
it reimposes the limits known to have been placed on the data by the
optics used to obtain it. Single-pixel events are "impossible"
because, assuming that Nyquist has been satisfied, the smallest
"real" feature in the data should be at least 4 pixels wide (or 12-16
pixels in area, 50-100 voxels in volume), not one pixel.

Because the spatial frequency of a noise pulse singularity is at
least 4x higher than that of the highest spatial frequency that the
optical system is capable of having transmitted, the offending value
can be tagged and then either replaced or averaged down. Indeed, some
EM-CCDs can now be set to detect and remove single-pixel noise based
on this recognition.

More generally however, the most reliable and robust method of
obtaining the most accurate pixel-intensity information from a series
of sequentially-obtained data sets is to deconvolve them in time as
well as space. This just means that we put into the process not just
the PSF (which sets the limits on possible spatial frequencies) but
also our knowledge that real changes in specimen brightness can only
occur so fast and not any faster.

George has postulated a series of intensity measurements from a
single pixel. Depending on the time delay between the measurements in
this series, we may (or may not) be justified in assuming that no
real change in the specimen could justify a sudden, 100x intensity
change that is only one scan-time in duration. Again, this allows us
to tag outliers and then dispose of them either by replacement or
averaging.

So much for what can now be conveniently done using computers.

Let us not forget that every effort should also be taken to reduce
the number of single-pixel anomalies present in the raw data to begin
with. With the PMT, this means keeping the photocathode cool and
small and monitoring its no-real-signal output over time (i.e., dark
count rate). Store a reference image of a single scan with all the
lasers and room lights turned off, and look for changes in its
general appearance as the weeks pass. More quantitative measures are
also wise. (And while you are at it, compare this zero-light result
with one obtained when the level of room illumination present is
similar to that which you use when actually collecting data. Stray
light is often a more serious problem than we we expect.)

When employing an EM-CCD, a similar no-signal image can be used to
assess changes in dark-count and coupling-induced charge over time.
These may slowly drift up over time (months) and are always very
sensitive to chip temperature.

I am less familiar with the anomalous, single-pixel behavior of sCMOS
cameras, but I would guess that, with the exception of hot pixels,
they are less common simply because charge amplification is not
involved and events associated with the emergence of a single, errant
photoelectron cannot be seen above the general read-noise level. As
hot-pixels tend to reoccur at the same exact location in the image,
they can be automatically identified and averaged out using data from
their 4 or 8 nearest neighbours.

In all these cases the most efficient way of removing single-pixel
anomalies from your final data is to take all the precautions needed
to prevent them from occurring.

Sorry for being so long-winded.

Jim Pawley

Jeremy Adler-4

Re: median filtering confocal microscope data at the instrument

In reply to this post by George McNamara

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

A detailed look at median filtering is quite surprising.

Assuming 5 images with an average photon count of 24 (mean intensity
23.5) and Poisson noise.

Standard deviation
average projection 2.21
median projection 2.64
so the median projection has a wider distribution than average
projection, but not too bad and the mean is almost identical.

The unexpected effect comes when very high values are included (Guy's
PMT noise) - simulated by adding 100 to one of the 5 images.
median standard deviation increases to 2.95, a trend you might expect
But more surprisingly the mean intensity of the median filtered image
goes up from 23.5 to 24.96, because all the outliers are above the
mean. This effect does not depend on the magnitude of the simulated
PMT noise - an offset of 200 has the same effect.
In this simulation the projected mean is very badly affected by the
PMT noise and the median filter Guy proposes is effective.

However the best result comes from rejecting the values that include
PMT noise, which is easy to detect and taking the mean of the
remaining values - the mean is fine and the SD then widens from 2.21
to 2.47.

On a wider note, a very good way of showing the levels of image noise
is to take pairs of images and display their similarity in a
scattergram (J. Microscopy 230(1),121-133).

If anyone can get this very simple and effective way of showing noise
into commercial acquisition sofware I will be very impressed -
implementation would not be difficult and the scattergram is very easy
to understand.

James Pawley

Re: median filtering confocal microscope data at the instrument

In reply to this post by Guy Cox-2

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

>*****
>To join, leave or search the confocal microscopy listserv, go to:
>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>*****
>
>Jim,
>
> OK, we are probably going to come to blows over this. I just
>trust the buffer of the Pacific Ocean between us. The term 'filter'
>applied to digital operations is a bit unfortunate. An optical
>filter removes light according to its specification. A digital, so
>called, filter does nothing of the sort. It processes pixels
>according to the values of other pixels. Deconvolution does EXACTLY
>the same thing - just with a more sophisticated algorithm.
>Fundamentally there is no difference. I really wish the term
>'filter' had never been used in the digital world.
>
> Guy

Well, not quite blows. And I agree that "filtering" and
"deconvolution" do have some similarities.

But I would like to point out the following:

The rationale for deconvolution is that, to the extent that one can
mathematically model the blurring effect of an imaging system as a
convolution, one should be able to reduce its blurring effect by
deconvolving the raw data. The one assumption is that the array of
point emitters in the specimen are blurred the the same PSF to
produce the blurred data that we detect.

In the case of deconvolving 3D microscope data, the main limitations
on this process are image noise (Poisson, as well as others) and the
possibility that the PSF is not perfectly known and may not remain
constant over the sampled volume.

Therefore, the assumptions that are put into any acceptable spatial
deconvolution system should be traceable to verifiable measurements
of, for instance, the optical and sampling parameters being used.
Deconvolving in time should be based on knowledge of how the
fluorescent signal is expected to change with time: the simplest
version being that it doesn't change during the acquisition period.

By contrast, the strongest support that I have heard for using, for
instance, a particular median filter is that it makes the image
somehow look better by suppressing occasional bright pixels. As far
as I know, one doesn't even have to input any sampling/PSF data
although the effects of such filters obviously vary with spatial
frequency. Therefore, we don't even know the size of the bright pixel
in real space.

Though the images that result from some filters may resemble those
produced by some deconvolution procedures, I feel that the former
inspires less confidence than the latter.

As a compromise I have suggested in the past that, as the process by
which the microscope "convolves" structural data is a 3D process, we
restrict the use of the term deconvolution to 3D data sets while
procedures that are applied to only a single plane of data (at any
one time) be called filters.

Cheers,

Jim Pawley

>-----Original Message-----
>From: Confocal Microscopy List
>[mailto:[hidden email]] On Behalf Of James Pawley
>Sent: Wednesday, 27 February 2013 3:45 AM
>To: [hidden email]
>Subject: Re: median filtering confocal microscope data at the instrument
>
>*****
>To join, leave or search the confocal microscopy listserv, go to:
>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>*****
>
>Hi all,
>
>It seems that we are discussing the best ways of eliminating the
>effects of what are sometimes called "single-pixel" noise events.
>Although it is fair to ask "What other kind of noise is there?", the
>term is often used to refer to pixels with recorded intensity values
>that are "unreasonably large" and seem to have nothing to do with the
>presence of dye molecules at a certain location in the specimen. Such
>values can come from a number of possible sources: cosmic rays pass
>through the photocathode every few seconds; alpha particles from
>radioactive elements in the PMT somewhat more often. If the PMT is
>used at high gain, as is often the case when looking at living
>specimens and signal levels must be kept low, single-photoelectron
>dark counts may produce fast pulses from the PMT or EM-CCD that
>approach the size of those representing signal in a "stained" pixel.
>As these signals seem obvious artifacts when viewed by eye, it would
>be convenient if they could be removed automatically.
>
>By definition, filters take things out. That is both their aim and
>their curse. One can argue for hours on whether or not the resulting
>data is better or not. In addition, filters are very fast. However,
>as computers get ever faster and cheaper one would expect that this
>advantage would become less important.
>
>In contrast to filters, deconvolution puts things in. Traditionally,
>it reimposes the limits known to have been placed on the data by the
>optics used to obtain it. Single-pixel events are "impossible"
>because, assuming that Nyquist has been satisfied, the smallest
>"real" feature in the data should be at least 4 pixels wide (or 12-16
>pixels in area, 50-100 voxels in volume), not one pixel.
>
>Because the spatial frequency of a noise pulse singularity is at
>least 4x higher than that of the highest spatial frequency that the
>optical system is capable of having transmitted, the offending value
>can be tagged and then either replaced or averaged down. Indeed, some
>EM-CCDs can now be set to detect and remove single-pixel noise based
>on this recognition.
>
>More generally however, the most reliable and robust method of
>obtaining the most accurate pixel-intensity information from a series
>of sequentially-obtained data sets is to deconvolve them in time as
>well as space. This just means that we put into the process not just
>the PSF (which sets the limits on possible spatial frequencies) but
>also our knowledge that real changes in specimen brightness can only
>occur so fast and not any faster.
>
>George has postulated a series of intensity measurements from a
>single pixel. Depending on the time delay between the measurements in
>this series, we may (or may not) be justified in assuming that no
>real change in the specimen could justify a sudden, 100x intensity
>change that is only one scan-time in duration. Again, this allows us
>to tag outliers and then dispose of them either by replacement or
>averaging.
>
>So much for what can now be conveniently done using computers.
>
>Let us not forget that every effort should also be taken to reduce
>the number of single-pixel anomalies present in the raw data to begin
>with. With the PMT, this means keeping the photocathode cool and
>small and monitoring its no-real-signal output over time (i.e., dark
>count rate). Store a reference image of a single scan with all the
>lasers and room lights turned off, and look for changes in its
>general appearance as the weeks pass. More quantitative measures are
>also wise. (And while you are at it, compare this zero-light result
>with one obtained when the level of room illumination present is
>similar to that which you use when actually collecting data. Stray
>light is often a more serious problem than we we expect.)
>
>When employing an EM-CCD, a similar no-signal image can be used to
>assess changes in dark-count and coupling-induced charge over time.
>These may slowly drift up over time (months) and are always very
>sensitive to chip temperature.
>
>I am less familiar with the anomalous, single-pixel behavior of sCMOS
>cameras, but I would guess that, with the exception of hot pixels,
>they are less common simply because charge amplification is not
>involved and events associated with the emergence of a single, errant
>photoelectron cannot be seen above the general read-noise level. As
>hot-pixels tend to reoccur at the same exact location in the image,
>they can be automatically identified and averaged out using data from
>their 4 or 8 nearest neighbours.
>
>In all these cases the most efficient way of removing single-pixel
>anomalies from your final data is to take all the precautions needed
>to prevent them from occurring.
>
>Sorry for being so long-winded.
>
>Jim Pawley
>
>
>>*****
>>To join, leave or search the confocal microscopy listserv, go to:
>>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> >*****
>>
>>Hi George,
>>
>>On Mon, 25 Feb 2013, George McNamara wrote:
>>
>>> So if a single pixel scanned five times has values of 40, 38, 42, 3800,
>>> 37, (not necessarily acquired in that order), you would prefer the
>>> arithmetic mean 791.4 (not that any of the vendors can give you the 0.4,
>>> and for anyone who is a fan of Kalman, the Kalman value would depend on
>>> which order the five were acquired in, rather than 40 (which is the
>>> "digital offset" I usually use on the Zeiss LSM710 I manage when
>>> operating in 12-bit mode).
>>
>>Please do not pretend that I said that. When I talked about "linear"
>>filters, I did not imply a simple averaging.
>>
>>> Pesonally, I would like to see the point scanning confocal microscope
> >> (and EMCCD software) vendors implement median and even more PMT and
>>> similar noisy data appropriate methods to provide the best possible data
>>> to my users and I (and as of April: my colleagues and I at MDACC,
>>> Houston).
>>
>>What data would be best recorded depends highly on the application. In the
>>general case, recording the values 40, 38, 42, 3800 and 37 in your above
>>example would be better than recording just "40". But recording more than
>>one value per pixel is often not practical.
>>
>>To reiterate: The Median filter *can* be the optimal filter. You should
>>just not go around and tell everybody that it *is* the optimal filter,
>>because it certainly is not. And in particular when you want to quantify
>>your data after acquisition, it is inappropriate to use the Median filter
>>(remember: a filter should not be applied just because the processed image
>>"looks good", but it should only be applied if it helps the analysis).
>>
>>That is all I said. (I certainly did not claim that you should always take
>>a simple arithmetic mean. I am not that stupid.)
>>
>>Ciao,
>>Johannes
>
>
>--
>James and Christine Pawley, 5446 Burley Place (PO Box 2348), Sechelt,
>BC, Canada, V0N3A0,
>Phone 604-885-0840, email <[hidden email]>
>NEW! NEW! AND DIFFERENT Cell (when I remember to turn it on!) 1-604-989-6146

Guy Cox-2

Re: median filtering confocal microscope data at the instrument

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Most digital filters (not median filters) are convolutions. Many deconvolution systems are filters (for example Wiener filters). Confocal and 4-pi images can be effectively deconvolved by an inverse filter. There is no basis for drawing a line here, it is a continuum. Nor can I support your idea that filters should only be applied on a 2D basis - a long time ago I showed that filters should be applied in as many dimensions as the dataset possessed. G.C. Cox and Colin Sheppard, 1999 Appropriate Image Processing for Confocal Microscopy. In: P.C. Cheng, P P Hwang, J L. Wu, G Wang & H Kim (eds) Focus on Multidimensional Microscopy. World Scientific Publishing, Singapore, New Jersey, London & Hong Kong. Volume 2, pp 42-54 ISBN 981-02-3992-0.

Applying filters plane by plane gave hugely worse results. Also, this paper showed that sampling slightly above Nyquist (3 pixels per resel) and then median filtering with a minimal (circular, or face contact only) kernel reduced noise very effectively while not impacting at all on resolution. So it cannot be throwing away information. Isn't that what we need?

Guy

>Jim,
>
> OK, we are probably going to come to blows over this. I just
>trust the buffer of the Pacific Ocean between us. The term 'filter'
>applied to digital operations is a bit unfortunate. An optical filter
>removes light according to its specification. A digital, so called,
>filter does nothing of the sort. It processes pixels according to the
>values of other pixels. Deconvolution does EXACTLY the same thing -
>just with a more sophisticated algorithm.
>Fundamentally there is no difference. I really wish the term 'filter'
>had never been used in the digital world.
>
> Guy

Well, not quite blows. And I agree that "filtering" and "deconvolution" do have some similarities.

But I would like to point out the following:

The rationale for deconvolution is that, to the extent that one can mathematically model the blurring effect of an imaging system as a convolution, one should be able to reduce its blurring effect by deconvolving the raw data. The one assumption is that the array of point emitters in the specimen are blurred the the same PSF to produce the blurred data that we detect.

In the case of deconvolving 3D microscope data, the main limitations on this process are image noise (Poisson, as well as others) and the possibility that the PSF is not perfectly known and may not remain constant over the sampled volume.

Therefore, the assumptions that are put into any acceptable spatial deconvolution system should be traceable to verifiable measurements of, for instance, the optical and sampling parameters being used.
Deconvolving in time should be based on knowledge of how the fluorescent signal is expected to change with time: the simplest version being that it doesn't change during the acquisition period.

By contrast, the strongest support that I have heard for using, for instance, a particular median filter is that it makes the image somehow look better by suppressing occasional bright pixels. As far as I know, one doesn't even have to input any sampling/PSF data although the effects of such filters obviously vary with spatial frequency. Therefore, we don't even know the size of the bright pixel in real space.

Though the images that result from some filters may resemble those produced by some deconvolution procedures, I feel that the former inspires less confidence than the latter.

As a compromise I have suggested in the past that, as the process by which the microscope "convolves" structural data is a 3D process, we restrict the use of the term deconvolution to 3D data sets while procedures that are applied to only a single plane of data (at any one time) be called filters.

Cheers,

Jim Pawley

>-----Original Message-----
>From: Confocal Microscopy List
>[mailto:[hidden email]] On Behalf Of James Pawley
>Sent: Wednesday, 27 February 2013 3:45 AM
>To: [hidden email]
>Subject: Re: median filtering confocal microscope data at the
>instrument
>
>*****
>To join, leave or search the confocal microscopy listserv, go to:
>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>*****
>
>Hi all,
>
>It seems that we are discussing the best ways of eliminating the
>effects of what are sometimes called "single-pixel" noise events.
>Although it is fair to ask "What other kind of noise is there?", the
>term is often used to refer to pixels with recorded intensity values
>that are "unreasonably large" and seem to have nothing to do with the
>presence of dye molecules at a certain location in the specimen. Such
>values can come from a number of possible sources: cosmic rays pass
>through the photocathode every few seconds; alpha particles from
>radioactive elements in the PMT somewhat more often. If the PMT is used
>at high gain, as is often the case when looking at living specimens and
>signal levels must be kept low, single-photoelectron dark counts may
>produce fast pulses from the PMT or EM-CCD that approach the size of
>those representing signal in a "stained" pixel.
>As these signals seem obvious artifacts when viewed by eye, it would be
>convenient if they could be removed automatically.
>
>By definition, filters take things out. That is both their aim and
>their curse. One can argue for hours on whether or not the resulting
>data is better or not. In addition, filters are very fast. However, as
>computers get ever faster and cheaper one would expect that this
>advantage would become less important.
>
>In contrast to filters, deconvolution puts things in. Traditionally, it
>reimposes the limits known to have been placed on the data by the
>optics used to obtain it. Single-pixel events are "impossible"
>because, assuming that Nyquist has been satisfied, the smallest "real"
>feature in the data should be at least 4 pixels wide (or 12-16 pixels
>in area, 50-100 voxels in volume), not one pixel.
>
>Because the spatial frequency of a noise pulse singularity is at least
>4x higher than that of the highest spatial frequency that the optical
>system is capable of having transmitted, the offending value can be
>tagged and then either replaced or averaged down. Indeed, some EM-CCDs
>can now be set to detect and remove single-pixel noise based on this
>recognition.
>
>More generally however, the most reliable and robust method of
>obtaining the most accurate pixel-intensity information from a series
>of sequentially-obtained data sets is to deconvolve them in time as
>well as space. This just means that we put into the process not just
>the PSF (which sets the limits on possible spatial frequencies) but
>also our knowledge that real changes in specimen brightness can only
>occur so fast and not any faster.
>
>George has postulated a series of intensity measurements from a single
>pixel. Depending on the time delay between the measurements in this
>series, we may (or may not) be justified in assuming that no real
>change in the specimen could justify a sudden, 100x intensity change
>that is only one scan-time in duration. Again, this allows us to tag
>outliers and then dispose of them either by replacement or averaging.
>
>So much for what can now be conveniently done using computers.
>
>Let us not forget that every effort should also be taken to reduce the
>number of single-pixel anomalies present in the raw data to begin with.
>With the PMT, this means keeping the photocathode cool and small and
>monitoring its no-real-signal output over time (i.e., dark count rate).
>Store a reference image of a single scan with all the lasers and room
>lights turned off, and look for changes in its general appearance as
>the weeks pass. More quantitative measures are also wise. (And while
>you are at it, compare this zero-light result with one obtained when
>the level of room illumination present is similar to that which you use
>when actually collecting data. Stray light is often a more serious
>problem than we we expect.)
>
>When employing an EM-CCD, a similar no-signal image can be used to
>assess changes in dark-count and coupling-induced charge over time.
>These may slowly drift up over time (months) and are always very
>sensitive to chip temperature.
>
>I am less familiar with the anomalous, single-pixel behavior of sCMOS
>cameras, but I would guess that, with the exception of hot pixels, they
>are less common simply because charge amplification is not involved and
>events associated with the emergence of a single, errant photoelectron
>cannot be seen above the general read-noise level. As hot-pixels tend
>to reoccur at the same exact location in the image, they can be
>automatically identified and averaged out using data from their 4 or 8
>nearest neighbours.
>
>In all these cases the most efficient way of removing single-pixel
>anomalies from your final data is to take all the precautions needed to
>prevent them from occurring.
>
>Sorry for being so long-winded.
>
>Jim Pawley
>
>
>>*****
>>To join, leave or search the confocal microscopy listserv, go to:
>>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> >*****
>>
>>Hi George,
>>
>>On Mon, 25 Feb 2013, George McNamara wrote:
>>
>>> So if a single pixel scanned five times has values of 40, 38, 42, 3800,
>>> 37, (not necessarily acquired in that order), you would prefer the
>>> arithmetic mean 791.4 (not that any of the vendors can give you the 0.4,
>>> and for anyone who is a fan of Kalman, the Kalman value would depend on
>>> which order the five were acquired in, rather than 40 (which is the
>>> "digital offset" I usually use on the Zeiss LSM710 I manage when
>>> operating in 12-bit mode).
>>
>>Please do not pretend that I said that. When I talked about "linear"
>>filters, I did not imply a simple averaging.
>>
>>> Pesonally, I would like to see the point scanning confocal
>>> microscope
> >> (and EMCCD software) vendors implement median and even more PMT
> and
>>> similar noisy data appropriate methods to provide the best possible data
>>> to my users and I (and as of April: my colleagues and I at MDACC,
>>> Houston).
>>
>>What data would be best recorded depends highly on the application. In
>>the general case, recording the values 40, 38, 42, 3800 and 37 in your
>>above example would be better than recording just "40". But recording
>>more than one value per pixel is often not practical.
>>
>>To reiterate: The Median filter *can* be the optimal filter. You
>>should just not go around and tell everybody that it *is* the optimal
>>filter, because it certainly is not. And in particular when you want
>>to quantify your data after acquisition, it is inappropriate to use
>>the Median filter
>>(remember: a filter should not be applied just because the processed
>>image "looks good", but it should only be applied if it helps the analysis).
>>
>>That is all I said. (I certainly did not claim that you should always
>>take a simple arithmetic mean. I am not that stupid.)
>>
>>Ciao,
>>Johannes
>
>
>--
>James and Christine Pawley, 5446 Burley Place (PO Box 2348), Sechelt,
>BC, Canada, V0N3A0, Phone 604-885-0840, email <[hidden email]> NEW!
>NEW! AND DIFFERENT Cell (when I remember to turn it on!) 1-604-989-6146

--
James and Christine Pawley, 5446 Burley Place (PO Box 2348), Sechelt, BC, Canada, V0N3A0, Phone 604-885-0840, email <[hidden email]> NEW! NEW! AND DIFFERENT Cell (when I remember to turn it on!) 1-604-989-6146

Johannes Schindelin

Re: median filtering confocal microscope data at the instrument

In reply to this post by Guy Cox-2

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

[Sorry, reposting with the subscribed mail address]

On Fri, 1 Mar 2013, Johannes Schindelin wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Hi Guy,
>
> On Wed, 27 Feb 2013, Guy Cox wrote:
>
> > The term 'filter' applied to digital operations is a bit unfortunate.
> > An optical filter removes light according to its specification. A
> > digital, so called, filter does nothing of the sort. It processes
> > pixels according to the values of other pixels. Deconvolution does
> > EXACTLY the same thing - just with a more sophisticated algorithm.
> > Fundamentally there is no difference. I really wish the term 'filter'
> > had never been used in the digital world.
>
> I still like to call it a "filter", and here is why: in digital images, we
> do not have photons, but we have information. Information is measured as
> entropy (the unit is "bits"). And no digital filter can increase the
> information. They can at most retain the same amount of information. But
> mostly they reduce information.
>
> So what about your deconvolution example?
>
> Let's go back first to the term "information" as per information theory.
> The amount of information in something like an image can be described as
> the average number of yes/no questions that have to be asked (given
> optimally efficient questioning) to describe it fully.
>
> Of course, this implies that we *already* know something, e.g. that it is
> a collection of pixels, in a certain geometric arrangement, the pixel
> values are in a certain range, etc. Without such a context, the
> information would be infinite and we would not be able to store it in a
> file.
>
> With deconvolution, we basically use additional knowledge about the image
> that is based on our assumption that the image formation happened a
> certain way, with a given point spread function. It is crucial to keep in
> mind that we reduce the amount of information in the original image using
> the knowledge about how the experiment works physically. It is even
> possible to put that information reduction into laymen's terms: we strip
> away the information about what the camera saw and retain only the
> information about the structures that gave rise to the acquired image.
>
> Sure, you could regenerate that image, but again you would need to use the
> knowledge about the optics; without that knowledge, the information is no
> longer in the deconvolved image. (And even with the knowledge, the
> reconstruction would be imperfect due to boundary effects, but that's
> beside the point.)
>
> Keep in mind that information always lives in a context. If you knew
> nothing about the bytes that make up this email, there would be no way to
> compress it. But since you know that it is written in English, using the
> ASCII encoding, you could compress it rather well. Even if you knew only
> that a human wrote it using a common computer, you could exploit the
> common knowledge that language is highly redundant, and compress it e.g.
> into a .zip file. (I like the compression example because it explains the
> unit "bits" and it illustrates the need for a context: .zip files compress
> rather poorly because the context "contains redundant and repetitive
> byte sequences" does not apply.)
>
> The same is happening with deconvolution: you have the context that you
> know a lot about the physics of image formation, and only that allows you
> to strip away the blurriness. Now, from the point of view of information
> theory, you strip away information. Which is good, because it is (mostly)
> information you do not care about.
>
> With this reasoning in mind, I hope that it is less offensive that I like
> the term "filter" in digital image processing.
>
> Ciao,
> Johannes
>

Lutz Schaefer

Re: median filtering confocal microscope data at the instrument

In reply to this post by George McNamara

Johannes
verbal interpretation of a mathematical matter is almost always incorrect. One has to be very careful not to be misunderstood. For example when you say that information gets stripped out of a widefield image is incorrect for deconvolution. In fact the inverse of a convolution will add intensities back into point objects. I cannot see how that reduces information content. My point is just to be careful when mathematical expressions become interpreted verbally. You likely loose information there too especially if the context isnt understood.

My 2 cents
Regards
Lutz

Sent from Samsung Mobile

-------- Original message --------
Subject: Re: median filtering confocal microscope data at the instrument
From: Johannes Schindelin <[hidden email]>
To: [hidden email]
CC:

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

[Sorry, reposting with the subscribed mail address]

On Fri, 1 Mar 2013, Johannes Schindelin wrote:

Johannes Schindelin

Re: median filtering confocal microscope data at the instrument

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi Lutz,

On Fri, 1 Mar 2013, Lutz wrote:

> verbal interpretation of a mathematical matter is almost always incorrect.

Given that I am a mathematician myself, one would hope that my command of
the mathematically-flavored language is good enough to deliver a correct
interpretation.

> One has to be very careful not to be misunderstood. For example when you
> say that information gets stripped out of a widefield image is incorrect
> for deconvolution. In fact the inverse of a convolution will add
> intensities back into point objects.

You say it will "add ... back". I think it would be better to leave the
"back" out since the information was never there!

Just imagine deconvolving a grayscale photograph of a bus. You quite
literally add information, and in this case it becomes quite obvious that
the information you put "back" was never there.

The same, however, is true for deconvolving a digital image acquired using
a CCD. The information about the point spread function was never in the
original image.

> I cannot see how that reduces information content.

But it does reduce the original information, that is what you use the
deconvolution for! After all, you do not want to see all that signal that
is caused by structures *and* point spread function. You only want to see
structures!

So what you do is to add more information (your knowledge of the physical
process) and reduce the combined information into a new, deconvolved image
that mostly contains the information you are interested in. Leaving out
the rest of the information you are no longer interested in.

> My point is just to be careful when mathematical expressions become
> interpreted verbally. You likely loose information there too especially
> if the context isnt understood.

Hence my attempt to clarify again.

Ciao,
Johannes

Guy Cox-2

Re: median filtering confocal microscope data at the instrument

In reply to this post by Lutz Schaefer

Lutz,

Are you really equating intensity with information? In a post about using accurate language to describe mathematical operations, this seems surprising.

Guy

-----Original Message-----
From: Confocal Microscopy List [mailto:[hidden email]] On Behalf Of Lutz
Sent: Saturday, 2 March 2013 6:22 AM
To: [hidden email]
Subject: Re: median filtering confocal microscope data at the instrument

Johannes
verbal interpretation of a mathematical matter is almost always incorrect. One has to be very careful not to be misunderstood. For example when you say that information gets stripped out of a widefield image is incorrect for deconvolution. In fact the inverse of a convolution will add intensities back into point objects. I cannot see how that reduces information content. My point is just to be careful when mathematical expressions become interpreted verbally. You likely loose information there too especially if the context isnt understood.

My 2 cents
Regards
Lutz

Sent from Samsung Mobile

-------- Original message --------
Subject: Re: median filtering confocal microscope data at the instrument
From: Johannes Schindelin <[hidden email]>
To: [hidden email]
CC:

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

[Sorry, reposting with the subscribed mail address]

On Fri, 1 Mar 2013, Johannes Schindelin wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Hi Guy,
>
> On Wed, 27 Feb 2013, Guy Cox wrote:
>
> > The term 'filter' applied to digital operations is a bit unfortunate.
> > An optical filter removes light according to its specification. A
> > digital, so called, filter does nothing of the sort. It processes
> > pixels according to the values of other pixels. Deconvolution does
> > EXACTLY the same thing - just with a more sophisticated algorithm.
> > Fundamentally there is no difference. I really wish the term 'filter'
> > had never been used in the digital world.
>
> I still like to call it a "filter", and here is why: in digital
> images, we do not have photons, but we have information. Information
> is measured as entropy (the unit is "bits"). And no digital filter can
> increase the information. They can at most retain the same amount of
> information. But mostly they reduce information.
>
> So what about your deconvolution example?
>
> Let's go back first to the term "information" as per information theory.
> The amount of information in something like an image can be described
> as the average number of yes/no questions that have to be asked (given
> optimally efficient questioning) to describe it fully.
>
> Of course, this implies that we *already* know something, e.g. that it
> is a collection of pixels, in a certain geometric arrangement, the
> pixel values are in a certain range, etc. Without such a context, the
> information would be infinite and we would not be able to store it in
> a file.
>
> With deconvolution, we basically use additional knowledge about the
> image that is based on our assumption that the image formation
> happened a certain way, with a given point spread function. It is
> crucial to keep in mind that we reduce the amount of information in
> the original image using the knowledge about how the experiment works
> physically. It is even possible to put that information reduction into
> laymen's terms: we strip away the information about what the camera
> saw and retain only the information about the structures that gave rise to the acquired image.
>
> Sure, you could regenerate that image, but again you would need to use
> the knowledge about the optics; without that knowledge, the
> information is no longer in the deconvolved image. (And even with the
> knowledge, the reconstruction would be imperfect due to boundary
> effects, but that's beside the point.)
>
> Keep in mind that information always lives in a context. If you knew
> nothing about the bytes that make up this email, there would be no way
> to compress it. But since you know that it is written in English,
> using the ASCII encoding, you could compress it rather well. Even if
> you knew only that a human wrote it using a common computer, you could
> exploit the common knowledge that language is highly redundant, and compress it e.g.
> into a .zip file. (I like the compression example because it explains
> the unit "bits" and it illustrates the need for a context: .zip files
> compress rather poorly because the context "contains redundant and
> repetitive byte sequences" does not apply.)
>
> The same is happening with deconvolution: you have the context that
> you know a lot about the physics of image formation, and only that
> allows you to strip away the blurriness. Now, from the point of view
> of information theory, you strip away information. Which is good,
> because it is (mostly) information you do not care about.
>
> With this reasoning in mind, I hope that it is less offensive that I
> like the term "filter" in digital image processing.
>
> Ciao,
> Johannes
>

George McNamara

Re: median filtering confocal microscope data at the instrument

In reply to this post by Lutz Schaefer

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi Lutz,

Thanks for posting to the listserv on this topic.

Are you going to take the information from this thread and incorporate
it into improvements in Zeiss ZEN confocal and widefield software (and
documentation, application notes, marketing bullet points, salespeoples
talking points, applications people training points)?

George

On 3/1/2013 2:21 PM, Lutz wrote:

> Johannes
> verbal interpretation of a mathematical matter is almost always incorrect. One has to be very careful not to be misunderstood. For example when you say that information gets stripped out of a widefield image is incorrect for deconvolution. In fact the inverse of a convolution will add intensities back into point objects. I cannot see how that reduces information content. My point is just to be careful when mathematical expressions become interpreted verbally. You likely loose information there too especially if the context isnt understood.
>
> My 2 cents
> Regards
> Lutz
>
> Sent from Samsung Mobile
>
> -------- Original message --------
> Subject: Re: median filtering confocal microscope data at the instrument
> From: Johannes Schindelin<[hidden email]>
> To: [hidden email]
> CC:
>
> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> [Sorry, reposting with the subscribed mail address]
>
> On Fri, 1 Mar 2013, Johannes Schindelin wrote:
>
>
>> *****
>> To join, leave or search the confocal microscopy listserv, go to:
>> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>> *****
>>
>> Hi Guy,
>>
>> On Wed, 27 Feb 2013, Guy Cox wrote:
>>
>>
>>> The term 'filter' applied to digital operations is a bit unfortunate.
>>> An optical filter removes light according to its specification. A
>>> digital, so called, filter does nothing of the sort. It processes
>>> pixels according to the values of other pixels. Deconvolution does
>>> EXACTLY the same thing - just with a more sophisticated algorithm.
>>> Fundamentally there is no difference. I really wish the term 'filter'
>>> had never been used in the digital world.
>>>
>> I still like to call it a "filter", and here is why: in digital images, we
>> do not have photons, but we have information. Information is measured as
>> entropy (the unit is "bits"). And no digital filter can increase the
>> information. They can at most retain the same amount of information. But
>> mostly they reduce information.
>>
>> So what about your deconvolution example?
>>
>> Let's go back first to the term "information" as per information theory.
>> The amount of information in something like an image can be described as
>> the average number of yes/no questions that have to be asked (given
>> optimally efficient questioning) to describe it fully.
>>
>> Of course, this implies that we *already* know something, e.g. that it is
>> a collection of pixels, in a certain geometric arrangement, the pixel
>> values are in a certain range, etc. Without such a context, the
>> information would be infinite and we would not be able to store it in a
>> file.
>>
>> With deconvolution, we basically use additional knowledge about the image
>> that is based on our assumption that the image formation happened a
>> certain way, with a given point spread function. It is crucial to keep in
>> mind that we reduce the amount of information in the original image using
>> the knowledge about how the experiment works physically. It is even
>> possible to put that information reduction into laymen's terms: we strip
>> away the information about what the camera saw and retain only the
>> information about the structures that gave rise to the acquired image.
>>
>> Sure, you could regenerate that image, but again you would need to use the
>> knowledge about the optics; without that knowledge, the information is no
>> longer in the deconvolved image. (And even with the knowledge, the
>> reconstruction would be imperfect due to boundary effects, but that's
>> beside the point.)
>>
>> Keep in mind that information always lives in a context. If you knew
>> nothing about the bytes that make up this email, there would be no way to
>> compress it. But since you know that it is written in English, using the
>> ASCII encoding, you could compress it rather well. Even if you knew only
>> that a human wrote it using a common computer, you could exploit the
>> common knowledge that language is highly redundant, and compress it e.g.
>> into a .zip file. (I like the compression example because it explains the
>> unit "bits" and it illustrates the need for a context: .zip files compress
>> rather poorly because the context "contains redundant and repetitive
>> byte sequences" does not apply.)
>>
>> The same is happening with deconvolution: you have the context that you
>> know a lot about the physics of image formation, and only that allows you
>> to strip away the blurriness. Now, from the point of view of information
>> theory, you strip away information. Which is good, because it is (mostly)
>> information you do not care about.
>>
>> With this reasoning in mind, I hope that it is less offensive that I like
>> the term "filter" in digital image processing.
>>
>> Ciao,
>> Johannes
>>
>>