Plustek OptikFilm 7200

[Kennedy McEwen]

Kennedy McEwen wrote:

You think you can measure something without dealing with uncertainties?

No, but you are proposing an uncertainty *AS* a measurement rather than
the limit of accuracy of a measurement.

But you wouldn't trust it, right?

If it produced measurements that were repeatable and correlated with the
established methods then I would trust it - but until you demonstrate
that, no, I don't believe it.

You might order that book though. One could say its central idea is
that in order to interpret an image you always need to estimate
'restore' the object from the available data.

But that is merely *one* step in the interpretation - and one which I
have already explained is only necessary with the limited test object
that you are using. What your "process" achieves is to transform the
spectral characteristics of the line or spot that you are using as your
test object into a flat spectrum. The edge method starts with a flat
spectrum, so the procedure that you have been wittering on about is
completely unnecessary as a preliminary step to calculating the MTF.

You are welcome to object against that, saying that one cannot measure
anything that way.

I haven't objected to it - in fact I have pointed out that it is a
necessary step in your method - however it does NOT overcome the
fundamentally undersampled nature of the measurement you are making.
Since you have undersampled data, no amount of deconvolution or further
processing will recover the information above the nyquist limit, nor
unravel the aliased information from the real.

I didn't say it was 'inadequate', merely that the choice to use an edge
is probably more driven by the need to have an easy to manufacture test
object than anything else.

But that isn't the reason it is used at all. Line measurements formed
the basis of all prior tests - whether the basic test of the viewing a
USAF-1951 chart or a more complex test pattern. The problem is two fold
- uncertainty introduced by undersampling and uncertainty introduced by
lack of resolved contrast or power. The edge test overcomes both of
these. The edge delivers the maximum transient power to the system
under test whilst the slope permits the edge to be oversampled by a
ratio determined by its gradient, thus eliminating both of the major
limitations of a line test.

As soon as it is easy to obtain bright point or line objects the
standard tests can indeed be updated.

Strange that the historical test methodology has been exactly the
opposite of what you have suggested - edge measurement has replaced line
measurement, not vice versa. This is yet another flaw in your plan -
the measurement is one of the scanner, and that includes its own
illumination source. Further, once it is bright enough to cause
saturation of the CCD in the centre of the line or point then there is
no benefit of it being any brighter - and that can already be achieved
with ease.