Posted by
Gerhard Holst on
URL: http://confocal-microscopy-list.275.s1.nabble.com/sCMOS-salt-n-pepper-issues-tp7583323p7583329.html
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopyPost images on
http://www.imgur.com and include the link in your posting.
*****
Hi Neil,
You did proper measurements and your findings are valid for sCMOS image sensors. It is very similar to the improvements in CCD image sensors before. Whenever the general noise behavior had been significantly improved, just in the lower noise signal additional noise phenomena became visible, that were not visible before, because they were hidden in the overall larger noise signal. It is definitively a image sensor issue, which can be seen in cameras of all the manufacturers.
The scientific CMOS have become very sensitive, very low noise, but still they are CMOS image sensors, with dark current, with traps in the silicon, with electrical connections at the surface. Yes, both the "higher" noise pixel in the dark as well as the "darker" pixel in the bright are part of the "blinker" and higher noise pixels of sCMOS. Sometimes, if the noise histograms are given with a logarithmic scale, you will see that there is a significant tail towards the larger noise pixels, which might be improved in the future by Fairchild by changes in the production process, which is a complex task due to the many knobs that could be turned.
In the dark, the amount of these guys will increase a little if the exposure time becomes very long, above 1 s. There you will see an increased offset value in some pixels, if you do an average measurement of a large stack of dark images, you will be able to see and detect the position of these pixels, which have not been put into the "hot pixel list" of the manufacturer. You could store their locations, and treat them either as defect pixels, try to use some kind of a weighted "salt-and-pepper" filtering or you do your own offset calibration in post processing. But these positions will not change, like the clock induced charge signals in emCCD for example.
Due to the high gain, keep in mind, that there only few electrons involved.
In the bright, there are some gain issues with some of the higher noise pixels. Again you can detect them, store the location and treat them as defect pixel or create your own additional gain calibration. Also here I would use a larger stack of homogenously illuminated medium bright images. You would need the averaging to exclude the photon noise.
In general, the sCMOS has nice performance parameters, but in my opinion it is not optimum for long exposures. Its is better for higher frame rates and shorter exposure times, where the dark current and its noise influence can be neglected.
with best regards,
Gerhard
_______________________________
Dr. Gerhard Holst
Science & Research
PCO AG
Donaupark 11
93309 Kelheim, Germany
fon +49 (0)9441 2005 36
fax +49 (0)9441 2005 20
mob +49 (0)172 711 6049
[hidden email]
www.pco.de
-----Ursprüngliche Nachricht-----
Von: Confocal Microscopy List [mailto:
[hidden email]] Im Auftrag von Neil Anthony
Gesendet: Mittwoch, 21. Januar 2015 04:24
An:
[hidden email]
Betreff: sCMOS salt'n'pepper issues
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopyPost images on
http://www.imgur.com and include the link in your posting.
*****
Hi all, I hope the
science/optics/hardware/software/students/post-docs/reviewers/managers/next-door-neighbors
and FedEx are treating you well.
We have a couple of issues with a recently installed Hamamatsu Flash 4.0
sCMOS camera and I was wondering if anybody has had any similar
experiences, both in terms of feedback on what is expected from sCMOS
hardware and what can be done to rectify any issues.
Here's what we're seeing:
Using a Hamamatsu Flash 4.0 camera link sCMOS (firmware v3.00A) on a
spinning disk and acquiring data using Perkin Elmer Volocity 6.3 we're
having three issues regarding hot and cold pixels. So far we have tried
4 different sCMOS cameras (kindly supplied/tested by Hamamatsu and
Perkin Elmer), to double check the subtleties, but the results are quite
similar
On the first camera we saw a bright cluster of 4 pixels that are between
2 and 10x greater than the surrounding pixels across nearly all imaging
conditions (not seen in the other three cameras). Are these bright
pixel clusters something that others have seen much? This issue is seen
in both Volocity and ExCap/HDImage, but was not seen on the test sCMOS
camera used to cross check. I also see 'ghost' clusters that look
similar but only seem to last for one frame... Again, is that something
seen in sCMOS cameras in general?
Salt:
In images with low signal and longer exposure times we see speckled
bight pixels using both Volocity and ExCap/HDImage, with an intensity
approx twice that of the background signal. Exposure times ~secs; low
signal <1000 counts per pixel. These speckled hot spots were seen with
all sCMOS cameras to some degree.
Pepper:
In almost the inverse situation, where we have higher signals with lower
exposure times were also seen on all sCMOS cameras (the first camera
with the clear hotspot had very bad 'peppering' compared to the others).
I know that the pco cameras used on the Deltavision OMX have reference
images applied to reduce the affects of these artifacts, and I was
wondering if that's something that can be applied post-acquisition. Any
thoughts and/or feedback on what's normal for an sCMOS, and how to get
the best out of these cameras would be much appreciated. sCMOS cameras
certainly are fast and have a great signal to noise, so am I expecting
too much from this generation of hardware?
Thanks
Neil
A couple of images to compare salt'n'pepper pixels seen on three cameras
are here:
https://drive.google.com/file/d/0BzfJAGNfrgieQ21abmVPbFBqY0k/view?usp=sharinghttps://drive.google.com/file/d/0BzfJAGNfrgieVjViTWNoeExpdDQ/view?usp=sharing