Real resolution of scanner (USAF 1951)

[gmikol]

EDIT: Looks like polls are currently broken here on DPUG. Please just add a reply to this thread if you want to participate. Thanks!! --Greg

Hi All--

A common method in testing the resolution of scanners is the USAF 1951 test target.

Here is a sample of one:

If you're not familiar with this test chart, each pattern is described by its group number, and element number. This picture is of Group 4 (on the left and bottom), and Group 5, on the right. Each group is broken into 6 elements. The element number is next to each set of bar patterns. Horizontal bars measure vertical resolution, vertical bars measure horizontal resolution.

In reading the chart, you determine the smallest pattern that is still resolved. In order to convert this to line pairs/millimeter (LP/mm), you use the formula:

2^(G+[E-1/6])

e.g. Group 4, element 6 is 2^(4.833333) = 28.5 lp/mm

Reading the chart is subject to some amount of subjective interpretation and bias. For that reason, I'm not telling what resolution this was scanned at or what scanner was used.

You may be able to tell that this scanner has a difference in horizontal and vertical resolution. What this poll is about, is to determine a "consensus" value for horizontal resolution (vertical bars). I'm trying to equate a perceptual value to an MTF% derived from a slant-edge test on the same scanner. If I can do that, then I can take the observer bias out of this for some upcoming scanner testing I hope to do.

I appreciate your participation in this poll.

--Greg

 

[Surf]

the last resolved set of VERTICAL bars - Group 6, Element 1

It's good idea to overprint numbers of groups and elements as they can not be visible to everyone.

 

[Les Sarile]

It's good idea to overprint numbers of groups and elements as they can not be visible to everyone.

You can use the Coolscan as a reference -> Scandig Nikon Super Coolscan 5000ED USAF 1951

 

[Surf]

You can use the Coolscan as a reference -> Scandig Nikon Super Coolscan 5000ED USAF 1951

Thanks. I can use my scanner instead.

 

[gmikol]

I can't believe this thread has been viewed 100 times and only one person "voted".

I'll add my personal view: Group 5, Element 5 is the last clearly resolved pattern of vertical bars.

Anyone else?

--Greg

 

[gmikol]

At Surf's suggestion, I added labels to the graphic. Here's the new image (can't edit my original post anymore).

--Greg

 

[Eric Rose]

Well I guess it's a matter of what you define as "clearly defined/resolved". As far as I'm concerned the last usable pair is half way between 5/2 and 5/3. Beyond that it's just fuzz.

 

[gmikol]

Eric...just do be clear, we're talking about the vertically-oriented bars?

But you make my point, exactly. There's a lot of interpretation that goes into reading a test chart like this. There is a generally-accepted value of 10% MTF for measuring resolution (it may even be in the ISO standard). But that 10% MTF doesn't match well with how I read a chart like the one I've included.

So while I doubt I'd ever get enough people to have a truly scientific sample, I was hoping to get enough opinions to correlate those results with some other MTF value. Perhaps it's just a fool's errand, perhaps not.

--Greg

 

[Les Sarile]

There are two resolutions and both have to be identified.

Vertical 6, 1
Horizontal 5, 4​

 

[gmikol]

There are two resolutions and both have to be identified.

Vertical 6, 1
Horizontal 5, 4​

Yes, for the puposes of fully describing the resolution of a given scanner, both are necessary.

Right now, I'm just trying establish a relationship between our collective perception, and the measured MTF. I've chosen to use the horizontal resolution for this purpose. That relationship should be the same regardless of which direction it is measured in. (i.e. USAF horizontal Rez versus horizontal MTF, vertical res vs. vertical MTF.)

Thanks for voting.

--Greg

 

[gmikol]

Just to add a little more info for people reading this thread who might not be as familiar with this stuff as I or others might be...

To get from lp/mm to pixels per inch (ppi):
PPI = lp/mm * 2 * 25.4

Why? You need at least 2 pixels to form a line and a space, and 25.4 is for converting millimeters to inches.

Group 5, Element 1 32.0 lp/mm 1625 PPI
Group 5, Element 2 35.9 lp/mm 1824 PPI
Group 5, Element 3 40.3 lp/mm 2048 PPI
Group 5, Element 4 45.3 lp/mm 2299 PPI
Group 5, Element 5 50.8 lp/mm 2580 PPI
Group 5, Element 6 57.0 lp/mm 2896 PPI
Group 6, Element 1 64.0 lp/mm 3250 PPI

--Greg

 

[gmikol]

Just giving this a bump to see if we can get a little more participation.

--Greg

 

[L Gebhardt]

6,1

 

[Halka]

I'll agree with Eric, with 5,3 being pretty much the last useful element.

Just out of curiosity, what scanner is this? Feel free to PM if you don't want to introduce any bias