I am currently testing the image quality performance of an optical system consisting of a telecentric objective lens and a scientific CMOS camera. To measure the image quality I have been taking images of a 1951 USAF resolution test target. I capture images continuously as the camera is subjected to a series of environmental tests including operation in a vacuum. I would be interested to know if anyone can recommend suitable software with which I can analyse my images accurately and reliably to determine any changes in image quality that occur during operation of the camera in a vacuum. An example image is shown in the attachment.
The lens being tested was discussed in one of my previous questions:
https://www.researchgate.net/post/Does_anyone_know_if_a_commercial_off_the_shelf_camera_lens_will_work_in_a_vacuum
Well considering that the image is in a sense proportional to the square of the intensity, it is difficult for me to believe a possibility extracting the actual MTF out of the image only... all phase information, phase reversals is missing then.
You may fill some gaps by comparing ideal vs. actual image, and fill other gaps with more or less strong assumptions.
as for SW packages, you may try e.g.:
http://elim.physik.uni-ulm.de/?page_id=1298
http://www.quickmtf.com/slantededge.html
http://www.iasj.net/iasj?func=fulltext&aId=74272
There are some freewares but I guess they are not powerful enough (e.g. Image J and sion Image). So, try to find the instruction book of Olysia M3 or similar software. Then you would be able to choose the best one.
I work with XRD Microscopy and am very interested in understanding this subject. Please send me the link or the complete image of the 1951 USAF resolution test target to [email protected]
https://www.researchgate.net/file.PostFileLoader.html?id=531b133fd4c118cb608b4646&key=60b7d531b133ee5a64
Thanks for your help.
There is a Matlab tool from one of the California universities. A quick look didn't find that, but I found this:
http://www.mathworks.com/matlabcentral/fileexchange/28631-slant-edge-script
This guy also has some Matlab code.
http://www.normankoren.com/Tutorials/MTF.html
I agree with Ralph that Matlab is a solid route. If have Igor or Origin they should have the functionality you are looking, fro as well. (Though Igor can get program instensive)
The easier way for me is to use ImageJ: http://imagej.nih.gov/ij/ and the MTF plug in develop by Carles Mitja: http://carlesmitja.net/2011/12/16/new-mtf-plug-in/
It's a mater of minutes to get use to the software and then is really fast to work.
Please keep in mind that any software (incl. the ones mentioned above) will require exact knowledge of the image you use to extract an MTF; if you have a specific res. target it might just be easier to write your own version.
The UASF resolution target suitability for extracting an MTF seems very limited to me, as it contains not very many different line spacings / spatial frequency components.
Better would be to use a target with a variable lines space grating or pitched grating. A simple line profile through such a grating already gives a good representation of the transfer modulation.
If you want to utilize the MTF only to get an estimate of the highest resolution achievable, then a quicker alternative way might be to merely compute the absolute squared of the Fourier transform of the image and perform an azimuthal average of it; this is called a RPSD (Radial Power Spectrum Density). Where resulting 1D graph (i.e. RPSD) will decay into the noise floor is where the ultimate resolution cut off of the image will be. See attached image for an example of a RPSD from an image of a Siemens star test pattern.
BTW - Computing the RPSD can be done for any image and does not require a resolution target with known feature sizes.
I suggest you to use the MATLAB Image toolbox. Most in general, you can perform a quality test by acquiring an image and subsequently to perform the sum and the subtraction of the successive images acquired. Now you can individuate the variation in intensity of the pixels in the dark region (with the sum operation) and in the light region (with the subtraction operation). Be careful to work with images in double format, to avoid cutting the intensity to 0 or 255.
You can also try this:
http://rsb.info.nih.gov/ij/plugins/se-mtf/index.html
Perhaps the software QuickMTF works! You can google it for more information about it!
We have had good results with sfrmat which originated at the I3A. The capability in Imatest also originates from sfrmat.
http://losburns.com/imaging/software/SFRedge/index.htm
QuickMTF is one of the inexpensive tools you can get. Besides I agree that the best way in fact is to use someone's or write a Matlab/Python/Whatever script for this purpose, because it is not much more than a Fast Fourier Transformation.
An ImageJ plugin for MTF measurements based on Siemens star targets can be found on http://dipsystems.de/mtf
This thread is old, but I have run into some other resources. First of all, Burns of Kodak was at the forefront of the development of Slant Edge MTF and he has shared his code:
http://losburns.com/imaging/software/SFRedge/index.htm
There is a commercial quality freeware, sort of open source, MTF Mapper:
https://sourceforge.net/projects/mtfmapper/
and there is a low costing program Quick MTF:
http://www.quickmtf.com/
The author of the MTF Mapper (Franz van den Bergh) has a blog where he explains the theory in depth:
http://mtfmapper.blogspot.com/
He compares his program with Imatest and Quick MTF
http://mtfmapper.blogspot.com/2016/04/mtf-mapper-vs-imatest-vs-quick-mtf.html
And there is a link of a number of images with synthesized images with known MTFs and representative noise including readout noise and shot noise in this blog:
http://mtfmapper.blogspot.com/2015/07/taking-on-imatest.html
I ran those in the automatic settings in Imatest, MTF Mapper, and Quick MTF and got vastly different answers. They ran from 0.78 at 0.25 cy/px to 0.31...On the same image!
You can also google ISO 12233 and find a ton of info.
Quick MTF, I think, has a tool for USAF 1951, but it might be another program.
Ralph Reinhold is correct. Furthermore, the ImageJ plugin "SE MTF 2xNyquist" presents answers that have much disparity between expectation and actual results when using well-characterized lenses and focal plane arrays. This disparity has also been noted by others between examples from Imatest and the ImageJ plugin.
- Badges
- Science topic
- Similar topics
- Correction