Re: calibration of color cameras

On Apr 6, 1:28 pm, heini <buerg...@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx>
wrote:
Thanks for the many answers!

Let's go through it.
@ Kevin:I like the Idea with the color checker card. Any Idea how to
ensure proper illumination on the ***? Neon light can have very
different spectra, but not "daylight" Should I try 12V halogen spots
at 10.6 V?

@  Bob: Yes, the cameras are from the same manufacturer. But a simple
white balance is not enough, we still have major differences in the
images.
@ Richard: Yes, we thought about taking simple flurescence filters,
too. Problem with LED-illuminated Zeiss is that most filters are too
thick to be put in the light-path. So We would have to put them on top
of the slides.
From fluorescent standards I know that the thickness of the
"specimen" (Slide/Coverglass/Filter-Combination) influences the
result. Does this apply for Brightfield standards, too?

LED-illumination also explains that we don't have a 3200K setting that
makes sense, we simply use the same voltage for all instruments.
@Bob: We need this for the 63x only, so this problem should be OK.

@ all: If we want to check the color cameras at customer sites: Any
suggestions how many colors to use and which colors?
How does saturation influence the result? Should i use ND filters by
default?

Thanks for your valuable input!
Heini

On Apr 4, 7:09 pm, Bob Chiovetti <rchiove...@xxxxxxxxx> wrote:



On Apr 3, 12:55 pm, heini <buerg...@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx>
wrote:

Hello,
we have some color cameras on our microscopes and somehow we never get
identical results with them.
We are analyzing blood smears and good color reproduceability is
absolutely necessary for automated image analysis.
So we started to look for color calibration tools -- and found none so
far!
OK, there are sheets one can photograph (with indirect light), but no
calibration slides for transmitted light we could use in the
microscopes without taking the cameras off.
Or, am i wrong?
We would love slides with more that 3 colors on them and known RGB or
whatever spectrum.
Maybe someone in this group knows where to find calibration slides for
color cameras or has a better idea?

Thanks

Heini

Heini,

Are your cameras all of the same type?  It's difficult to say what to
do without more details, but lots of the cameras that are used for
microscopes have an "auto white balance" function, or something
similar, that usually works quite well.  This is found somewhere in
the software that's driving the camera.

In the simplest case, you move the slide away from the specimen so
that you are only looking at the transmitted light going through the
blank slide.  At this point you bring up the "auto white balance" or
"white balance" function and activate it.  The camera takes a picture
of this background, analyzes it and corrects it so the image is white
in color.  After that, the correction is applied to the pictures you
take of the specimen.

Some software packages have a "color selection" tool or an eyedropper
that you can point to an area that you want to make white.  You click
on this area, and the same thing happens: that area is made white, and
the correction is applied to the whole field of view.

Other cameras, especially consumer digital cameras, do this all by
software.  They analyze and correct the color from the entire field of
view.  Usually you have to set this up by telling the software what
kind of illumination you're using (incandescent, fluorescent,
daylight, etc.)  Some trial and error may be needed to get the correct
color, or something quite close to what you're seeing through the
microscope.  With these types of cameras, you can often set up and use
"custom" settings that will have the necessary color correction in
them.

A note of caution: it depends on the camera and the software, but it's
not unusual to have to correct the color or change the settings
whenever you change the magnification or adjust the illumination.  A
different objective or a change in the voltage to the bulb can change
the color of the illumination.  So you might need a different setting
if you go from a 10x to a 40x objective or if you decrease or increase
the brightness knob.

In the latter case I'd recommend that you use neutral density filters
to cut down or increase the brightness of the illumination rather than
constantly increasing and decreasing the voltage to the lamp in the
microscope.

Good luck, hope this helps!

Cheers,

Bob Chiovetti
swpinet dot com- Hide quoted text -

- Show quoted text -

With LED illumination, voltage is irrelevant.

Color cards won't help in any way, shape, or form with calibration to
an image analysis program.

Each instrument MUST, if you're doing any kind of analysis which
measures by color (which is nothing but 3 grey values), be calibrated
separately to the system you're using. It doesn't matter what they
look like on your monitor, what matters is what is being
measured.

If you're using an automated image analysis program, and if that
program measures color, then it should, if it's worth the money you
spent on it, have the ability to calibrate colors and/or
intensities. In this manner, you will take a picture at the
specified standard, and tell the system "this is the standard". At
that point, every picture will then be applied to the standard, and
measured.

What it "looks like" to you isn't relevant once it's calibrated, it
can look like the Jolly Green Giant, but as long as it's calibrated to
the software, and that is at a known, repeatable setting, then you are
collecting valid data. If you're NOT calibrating the software, and
it's not to a repeatable setting, then your data, no matter how
voluminous, is garbage.
.

Quantcast