Tuesday, March 25, 2014

How to Measure MTF, Part 2

Albert Theuwissen continues his MTF measurement series. The second part talks about the lens F-number affecting the sensor's MTF.


  1. Harvey (Herve) HornungMarch 27, 2014 at 4:57 PM

    There's just so more MORE to say about MTF measurements.
    Why do it with a (sine) Siemens star in the first place? How can you measure frequencies up to Nyquist at the RAW level?

    1. Albert TheuwissenMarch 28, 2014 at 1:53 PM

      Herve, I am using in this measurement a Siemens star because it is so easy working with it. It is just a matter of grabbing images and then further processing the data. I still have in mind also to do the measurements by means of the "knife edge", but that is for later. I do have one little, tiny issue : lack of TIME !
      But I am happy to accept your comments/additions to the blog. So feel free to add.

    2. Actually, there is a distinct advantage to using the Siemens star. The MTF that is typically shown is a 1 dimensional function that is the norm of a slice through the Fourier transform of the PSF. The problem here is that the PSF is not necessarily isotropic (and, therefore, the transfer function is also not isotropic).

      We know that PSF introduced by optical systems can be anisotropic (http://en.wikipedia.org/wiki/Zernike_polynomials). Siemens star can give an indication of some of the anisotropy in the PSF.

      There's a lot of interest in "computational imaging" recently, and the computation that is done after capturing can introduce deleterious effects on the actual signal (albeit producing a more aesthetically pleasing output). These computational effects can introduce nonlinear artifacts. Additionally, some systems may try to cut corner by not computing every frame exactly the same way based on hysteresis.

      I wonder if there is some sort of consortium on this stuff, but someone should definitely explore using more sophisticated non-parametric system identification techniques to characterize exactly what the entire imaging system is doing to the ground truth.

    3. Harvey (Herve) HornungMarch 29, 2014 at 2:38 AM

      Albert: it is easy because you have the chart and measurement software handy! For most people slanted edge is simpler because you can get the software only and print a reasonable chart from a simple printer.
      If you're interested, I can write some content that you can publish on your blog.

      thang: I doubt the Siemens star measurement can provide the PSF, it can provide MTF through several directions, but other methods can also do that.

  2. Albert TheuwissenMarch 28, 2014 at 7:29 PM

    Please notice that up to now for the MTF measurements a MONCHROME sensor/camera has been used. Colour will follow later.


All comments are moderated to avoid spam and personal attacks.