Search the CONFOCAL archive at
http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
I am pasting in a script file from an old Photometrics AT200 system that
shows the steps they used for the calculation. You will have to read
between the lines a bit - I added some comments as I was trying to
follow it once. Hope this is useful to someone....
Dale
* Photometrics AT200 system ==========
echo on
* Camera noise is measured using
* image1 = Bias_1 - Bias_2
* noise = stddev_1 * gain * sqrt(2.0) / 2.0
*
echo off
* do it quietly
icreate bias1 ccd
* allocates the memory and creates window for image
icreate bias2 ccd
* allocates the memory and creates window for image
ci bias2
* make bias2 be the current image
bias
* gets a bias (exp time = 0) readout of camera
ci bias1
bias
i- bias2
* subtracts bias2 from the current image (bias1)
* bias1 = bias1 - bias2
statistics -q
* computes statistics ("quietly", including std dev) of ci
noise
* calculates the noise: noise = stddev_1 * gain * sqrt(2.0) / 2.0
iremove bias1
iremove bias2
* ==============
====================Notes===============================
AT200 Camera Gain
The Gain setting determines the amplification of the digital signal
generated by charge on the CCD. The camera electronics unit converts
electrons on the CCD into digital intensity levels according to a
hardware-set base gain value. The base gain value is multiplied by a
factor determined by the software gain. If the gain value is 0, the
system will operate with no additional gain, currently, any positive
gain value will increase the signal generated by each electron by a
factor of four. This can be used to improve the definition of spectral
images, but the A/D converter will reach saturation sooner, effectively
decreasing the functional range of the CCD.
========================================================
John Oreopoulos wrote:
> Search the CONFOCAL archive at
> http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal Yes Jim, you're
> right. What I mentioned was only a part of the process described in the
> book. But I do remember you need at least one blank image for the
> calculation. As was pointed out earlier, you also need to know the
> conversion from ADUs to photoelectrons for your camera as well. The book
> recommends looking up this factor from the spec sheet, but there must be
> a way to measure this too.
>
> John
>
>
> On 11-Jul-07, at 11:40 AM, James Pawley wrote:
>
>> Search the CONFOCAL archive at
>> http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
>>> Search the CONFOCAL archive at
>>> http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal Dear Winnok,
>>>
>>> There is an excellent discussion about SNR and how to actually
>>> measure it with real images in the book "Fundamentals of Light
>>> Microscopy and Electronic Imaging" by Douglas B. Murphy starting on
>>> page 299. I think the answer you're looking for is there, and it do
>>> the measurement properly, you need to collect an image with the
>>> shutter to the camera closed - this becomes your reference dark image
>>> which you can measure and calculate the noise of the imaging system from.
>>>
>>>
>>> John Oreopoulos, BSc,
>>> PhD Candidate
>>> University of Toronto
>>> Institute For Biomaterials and Biomedical Engineering
>>> Centre For Studies in Molecular Imaging
>>>
>>> Tel: W:416-946-5022
>>>
>>
>>
>> Hi John,
>>
>> I haven't read the book you mention but it is hard for me to imagine
>> how one takes account of Poisson Noise using only a black image. At
>> best, one might get read noise. But this would require subtracting 2
>> black images (to cancel out fixed-pattern noise).
>>
>> To get Poisson noise, one must know the number of photons represented
>> by the recorded digital signal. On a CCD this means we have to know
>> the gain-setting (photons/ALU) of the camera.
>>
>> On the other hand, one could get some idea of the noise by taking two
>> "identical" images, calculating an average value and a difference
>> value for each pixel, and make a plot of the difference VS the
>> average. The scatter in this plot should increase with average value
>> in a way that represents Poisson noise. At low average values, it
>> should approximate the read noise.
>>
>> The only way I can image getting SNR info from one image is if you
>> first "deconvolve" (in 2D, Gaussian filter?) the data. Assuming that
>> it was Nyquist sampled, this will give a much less noisy "image" which
>> can be used as a reference. Subtract it from the original data and
>> plot the pixel-by-pixel difference against the signal intensity in the
>> deconvolved result. This would give a hand-waving approximation of SNR.
>>
>> Cheers,
>>
>> Jim P.
>>
>>>
>>> On 11-Jul-07, at 7:34 AM, winnok ugent wrote:
>>>
>>>> Search the CONFOCAL archive at
>>>> http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
>>>>
>>>> Dear all
>>>>
>>>> In order to test robustness of an image processing algorithm, I want
>>>> to estimate the signal to noise ratio of my images. It concerns a
>>>> very rough approximation since I merely want to take Poisson noise
>>>> into account, hence making abstraction of detector noise, dark noise
>>>> etc. (at least for now)
>>>> I want to be able to compare with synthetic images (originally
>>>> without noise) in which I gradually raise the Poisson noise. When
>>>> considering the original noiseless image as O, the current image as
>>>> C and i the voxel number I approximate the signal-noise ratio in
>>>> these synthetic images using (Manders et al., 1993):
>>>>
>>>> SNR=20log(sqrt(sum(Ci)"/sum(Ci-Oi)"))
>>>>
>>>> Since there is no original image for real images, this formula
>>>> doesn't apply. I have found several definitions of defining SNR, but
>>>> none of them seem to give values that correspond with what I
>>>> retrieve after using the first formula on the synthetic images. The
>>>> most common way seems to be using the standard deviation of a region
>>>> with 'no' signal to estimate noise and using the dynamic range or
>>>> maximum of the signal. But this seems rather arbitrary and dependent
>>>> on the chosen region. Is there a way of approximating the noise and
>>>> SNR of a single image - so not by recording the image several times
>>>> - without having to select a signal-free ROI and what would be the
>>>> most reliable definition?
>>>> Many thanks in advance.
>>>>
>>>> Kind regards,
>>>>
>>>>
>>>> winnok
>>>>
>>>>
>>>>
>>>>
>>>> ______
>>>> ir. Winnok De Vos
>>>> Research Assistant
>>>>
>>>> dep. Molecular Biotechnology
>>>> Faculty of Bioscience Engineering
>>>> Ghent University
>>>> Coupure links 653
>>>> 9000 Ghent
>>>> Belgium
>>>>
>>>> tel 0032.(0)9.264.59.71
>>>> fax 0032.(0)9.264.62.19
>>>> www.molecularbiotechnology.ugent.be
>>
>>
>> --
>> Jim Pawley (Summer address) c/o Postmaster, Egmont, BC, Canada,
>> V0N-1N0 604-883-2095, [log in to unmask] <mailto:[log in to unmask]>
>>
>
|
