Jackson K.M. Roland has shown that the slant edge MTF method is not influenced much by noise or the contrast ratio of the image. Google his write up titled "A Study of Slanted-Edge MTF Stability and Repeatability". I think that one implication of this work is that it should be possible to print slant image targets on an ordinary inkjet printer and photograph those with a digital camera to determine the resolution of the lens on the camera.
(...)
ImageJ is a free program image processing program that has an MTF method available. I have used it, but I don't remember if it was a plug in or a built-in application.
(...)
I'm not going to try to discuss the magic that allows one to beat the Nyquist limit, but basically I believe it has something to do with the fact that imaging of the slant edge basically makes it possible to (conceptually speaking) interpolate between the sampling points without introducing aliasing, at least up to a point. The phrase "phase offsets" often pops up in discussions of this.
One does have to make sure that all digital sharpening schemes are turned of everywhere in the system. I think some digital cameras automatically apply some sharpening, which could be a problem.
Thank you for the reference to JKM Roland's paper. Note that the SPIE publication is behind a paywall, but on the imatest site it is freely accessible.
ImageJ is standalone; except of course for acquisition (digicam or scanner)
There is no confict with the Nyquist limit: with the slanted-edge method, one can obtain arbitrary fine sampling when the slanted edge is almost, but not quite perpendicular to the scanning direction. In round numbers, a 15° misalignment (wrt to perpendicular to scan direction) gives 4x oversampling, 7.5° 8x oversampling, etc, as long as the edge is indeed straight. That is the key point of the slanted-edge method. The wording "phase effects" in the article makes it sound more mysterious than it is. The slanted edge provides multiple scans of the edge, with small incremental offsets; merging the multiple scans results in a single scan with fine sampling;
Digital sharpening in the camera's firmware is not a problem per se: consider the camera as a black box (even if in silver finish) and consider that you are characterizing the
system. Unless of course you want to characterize the lens rather than the lens+camera.