Posted by
Michael Schmid on
Feb 04, 2009; 3:20pm
URL: http://imagej.273.s1.nabble.com/Laser-Beam-Profile-Analysis-in-ImageJ-tp3693862p3693864.html
Hi,
a few thoughts about the background subtraction problem:
For your application, the built-in "Subtract Background" will
probably work best with the "sliding paraboloid" instead of the
default rolling ball method. Use a very high value of the radius.
You can select "create background" (don't subtract) and then use the
Image Calculator to subtract the background.
An alternative might be somehow selecting the surrounding not
irradiated by the laser beam and fitting a polynomial to this, using
the plugin on the documentation wiki:
http://imagejdocu.tudor.lu/doku.php?
id=plugin:filter:fit_polynomial:start
With order=0 you can get a constant value, with order=1 in x&y you
get a linear fit.
If you have a constant background and you want to automate it, but
you have no area that is always unilluminated, the 'mode' (peak) of
the histogram might be a good approximation for the background level.
Concerning the beam analysis itself:
The Moment Calculator should be perfect for Gaussian or near-Gaussian
beams.
For more top-hat beams, set the threshold to 10% and 90% of the
maximum and determine the area or size of fitting ellipse, size of
enclosing rectangle, whatever you want (Analyze>Set Measurements).
If you want to get accurate values for 10%/90% size, integral over
the beam, etc., the average value of the background must be zero.
This means that
(1) you need a 32-bit image (floating point; allowing for negative
values) and
(2) you must not use the built-in "subtract background" as the last
step. It puts the zero level below the noise of the background,
resulting in a nonzero average value of the background.
Michael
________________________________________________________________
On 30 Jan 2009, at 08:53, John Kershaw wrote:
> mleigh,
>
> I am doing some similar work with laser line beams in a field
> application
> measuring tree dimensions - so very variable and dynamic
> conditions. I've
> tested both red and green lasers and think I am going to use green
> - it is
> more visible outside. I have had some good success with using channel
> ratios: 2*red/(green+blue) or 2*green/(red+blue) to make a
> composite grey
> scale image. I use a maximum ratio, then scale everything between 0
> and
> 255. I have been using a number of line extraction masks to detect
> lines,
> measure height on the photo and length of the line - I am sure
> there are
> similar masks for circles.
>
> JAK (john kershaw)
>
> On Thu, Jan 29, 2009 at 11:01 AM, mleigh <
[hidden email]> wrote:
>
>> We are trying to use ImageJ for laser beam profile analysis.
>> Thus, we want
>> to accurately measure power in the beam, and use multiple methods to
>> calculate beam diameter. Also, we are analyzing field tests where
>> the
>> background changes from frame-to-frame due to clouds, sun
>> movement, and
>> other environmental effects.
>>
>> So far, I have not found any "laser beam profile" specific plug-
>> ins. Does
>> anyone on the list know of any that have been written? I have
>> found the
>> "Moment Calculator" which calculates the second moment, and thus
>> can be
>> turned into a "4 Sigma" beam width. We are also looking to
>> calculate the
>> 1/e^2 width when the distribution is summed in the x and y
>> directions, as
>> well as the "90/10" beam width.
>>
>> Background subtraction is also a problem for us as the built-in
>> "Rolling
>> ball" method does not work well for beam images. Other background
>> subtraction methods also do not work as well since we have a dynamic
>> background, and we want to preserve the pixel counts in the beam
>> area for
>> accurate measurement, and not just make a pretty picture. Right
>> now we are
>> taking two regions of interest per frame, and approximating the
>> background
>> as constant over the frame. However, we would like to fit a well-
>> behaved
>> function to the background, and subtract off the background on a
>> frame-by-frame basis for an entire stack. We would like to do it
>> sort of
>> like it's done in the "A_Posteriori" plug-in, but using a region of
>> interest, doing it on an entire stack, and subtracting the background
>> without renormalizing the output.
>>
>> If anyone has any ideas of ImageJ plug-ins that might implement
>> some of
>> these measurements, we would really appreciate it. ImageJ is so much
>> better
>> and faster than the Ophir/Spiricon BeamStar software we were
>> using. After
>> fighting BeamStar, we are very happy with ImageJ.
>>
>>
>>
>> --
>> View this message in context:
>>
http://n2.nabble.com/Laser-Beam-Profile-Analysis-in-ImageJ-
>> tp2239340p2239340.html
>> Sent from the ImageJ mailing list archive at Nabble.com.
>>