How much MTF is enough?

Phil Rhodes · February 16, 2019

I've been looking at documentation recently on comparative resolution figures. Inevitably, most of them eventually start talking about MTF.

Various people cite various percentages of MTF as a cutoff point for "effective resolution."

Has there ever been an MTF number that would be widely accepted as enough modulation to be considered a reasonable reproduction of an image - that is, a cutoff point for resolution? My impression is that there isn't any such widely-agreed metric for what constitutes resolution and what doesn't, perhaps because that would be an oversimplification.

P

David Mullen ASC · February 17, 2019

If you could explain MTF more clearly to the rest of us, you'd be doing everyone a favor!

Samuel Berger · February 17, 2019

I looked it up and it appears Phil means Modulation Transfer Function. So, not something I can help with, sadly.

Gregg MacPherson · February 17, 2019

I wish it was as easy as explaining the acronym :rolleyes: I always stumbled wondering what the minimum contrast is that the human eye can discern. Who's eye. We're all different. Some people have extreme acuity with seeing. There is probably a definition explaining that, somewhere, in very, very fine print, at very low contrast, so we can almost read it.

Phil Rhodes · February 18, 2019

Eh, sorry, my bad.

Shorthand definition:

When we're testing lenses, codecs, digital storage, systems, and so on, we often use charts with black and white stripes at various pitches and evaluate how much they're reduced in contrast by the system we're testing. As lenses, sensors etc approach their resolution limits, we start to notice that the black and white stripes start to merge into grey, effectively having their contrast reduced; we might say that at 1000 line-pairs per millimetre, contrast is reduced 50%. At some point the degree of contrast reduction in fine detail exceeds our personal tolerance and we consider that a resolution limit for the system.

More formal definition:

That 10%-at-1000 figure can be seen as a point on a graph which would show the frequency response of the system, like a spectrum analyser in audio.

Modulation transfer function is (formally) a subset of optical transfer function and is defined as the Fourier transform of the point spread function. This effectively means it's a frequency-domain expression of the behaviour of the optical system.

P

Aapo Lettinen · February 18, 2019

I think at the beginning of the digital era it was common to measure film mtf with approx 30% or 20% response and digital with 0%.

that is because the largest film grains mask the fine details sooner (finest details drawn by the smallest thus least sensitive grains) so there is no point to try to see the 0% details on film whereas with digital it is somewhat possible