Resolution test target for film scanner?

[jay_bees]

<p>I would like to find a resolution test target that will let me quantify and compare resolution on film scanners (need transparent not reflective target).</p>

<p>I am aware of the 1951 USAF glass slide resolution targets:<br>

http://www.edmundoptics.com/onlinecatalog/displayproduct.cfm?productID=1790<br>

<img src="http://www.edmundoptics.com/images/catalog/6033.gif" alt="6033.gif" width="284" height="182" align="middle" /><br>

but was wondering if there were any other options for resolution test targets available. It would be nice if I3A targets meeting the ISO 12233 standard were available in a transparent form for use in a film scanner.</p>

 

 

 

<img src="http://www.edmundoptics.com/images/catalog/2287.jpg" alt="" />

 

 

[jay_bees]

<p>Thanks.<br>

I'm actually interested in comparing various slide duplication techniques to scanners (such as Nikon 5000 and Imacon 949). I need a transmissive target that will let me quantify resolution.</p>

[jay_bees]

<p>Thanks Les.</p>

<p>Optimally, I would like to be able to quantify the resolution of the scanners, such as scanner A had a real resolution of 3300 ppi, scanner B 3700 ppi, and another at 4100 ppi, etc. Any targets for this use?</p>

[jay_bees]

<p>Great info Les. Thanks.</p>

[jay_bees]

<p>Les, a few questions if you do not mind.<br>

Did you print out that target yourself? What size was it printed at? What size was the full frame array of targets that you photographed?</p>

[kelly_flanigan1]

<p>CONTRAST radically effects test results with scanners and lenses:</p>

<p>If one tests a device with a 1:1000 contrast test negative; one might get results of = X line pairs per mm</p>

<p>If one tests a device with a 1:8 contrast test negative; one might get results of = 0.8 of X above line pairs per mm</p>

<p>If one tests a device with a 1:1000 contrast test negative; one might get results of =0.45 of X above line pairs per mm</p>

<p>Thus if one uses super high contrast targets; one gets high numbers. It is sort of like warming up your car driving it around; adding 50 psi to the tires; driving up to interstate topping of the gas tank and driving on a trip at 55mph non stop and getting fantastic gas mileage. You best cased the test and got numbers folks normally never get in real life.</p>

<p>What you really want to use is test targets that represent what YOU are trying to compare; so one has a more realistic test. Thus if your gig is 35mm Kodachromes; use some sample 35mm Kodachromes. If it is microfilm; use microfilm. If it is C41 35mm negatives; use some of our own C41 negatives.</p>

<p>In reality the public does not shoot 1:1000 test targets; they shoot low and medium contrast things. Using only 1:1000 test targets really gives one the limiting best case resolution of scanners 1,2,3, 4 etc. Scanner 2 and 3 might *test* the same numbers with 1:1000 references; but #3 model can be alot better when a 1:10 test target is used. The lay public ignores this and often cannot fathom a scanner has a transfer function of resolution that drops when the test target drops in contrast.</p>

<p>Here we got our first 35mm slide scanner back in 1989; ie 21 years ago. I use for testing some lens test panatomic-x negatives that were of 1:1000 USAF charts were a Konica Auto S2 at F5.6 recorded about 92 line pairs per mm; one of the best tests I have ever shot. I also use a similar set where I used USAF charts that were grey at 1:10 bars; and a Kodak reference glass slide that has vertical and horizontal progressively smaller bars.</p>

<p>What I do is I jsut test out each new flatbed; 35mm film scanner we buy with our own references. Thus if say our Canon FS4000 35mm film scanner pulls out alot of the details at 4000 dpi; and our latest 4800 or 6400 dpi flatbed pulls out say one half; I know that our flatbed is not really a 4800 or 6400 dpi device; maybe they only test in at 1800 to 2400dpi. I down sample the great drum or 4000 dpi film scanners image to appear to look like the ACME6400 flatebeds image. Thus experimentally I TEST for each devices *real world* specs. It really doesnt matter if another drinks the Koolaid and believes an Acme 10000 flatbed scanner is a 10,000 dpi device. In Engineering if I cannot measure it it is crowned marketing BS. Others/Lemings will jump in and say an Acme 10000 flatbed has an optical resolution of 10000; and thus I am in error.</p>

<p>If you scan for the public like I do; nobody brings in test targets to scan. They bring in shoe boxes of moldy gobs of Instamatic slides; faded 620 Kodacolors shot with box cameras; messes of 35mm slides; gobs of C41 35mm stuff. *ALOT* is not this best case stuff that taxes a scanner if it is the common Joe Six packs stuff.</p>

<p>I have some "old GIANT super sharp negatives to scan" is what a lady mentioned on the telephone last week. I am thinking 4x5" or 616/116 stuff. It turns out to be 6x6cm stuff; old 120/620 Kodacolors shot with a box camera. They are not even very sharp; our Epson 1200U flatbed from 10 years ago is a vast overkill; I scanned them at 1200 dpi; they really have only about 600 dpi worth of info since they are so overexposed; portraits shot too close with a fixed focus box camera. The background clutter is way sharper than the persons face.</p>

<p>You really want test targets that have mythical best case lab/bragging conditions; and some test targets with that are not the amateur/goosing to look good too; ie real world targets. Having some test targets with practical real world moderate contrasts like 1:10 or 1:5; 1:2 is what really makes scannera be more comparable; since it represents 99.999 percent of the publics "stuff". The publics stuff is not shot all at F5.6 with panatomic-X with Summicrons on granite blocks with timed lights. The public stuff is shot alot with iso 800 stuff; with kit zooms; with autofocus that grabbed the wrong kid in soccer.</p>

<p>Testing scanners for "resolution" is very old; it is old than photo.net It was being down before Photoshop or even digital cameras. Newcomers tend to fall into the best case trap; they use 1:1000 targets and crown their scanner a 2345 dpi device. Some of us who have the same scanner will look at these claims as being totally crazy/insane; they just make bloaded files which are useless.</p>

<p>In working on check scanners; we did use some best case references sometimes; but using actual checks written by humans is where the tire hits the road. One has weird ink colors that do not scan as well; one has faint lettering; pencil too.</p>

<p>The whole transfer function of a scanner varies by color too; faint 1:2 or 1:5 contast yellow or blue lines on whiten are not going to give the same resolutions as 1:1000 contrast black targets folks worship as gospel.</p>

<p>If you want goosed pumped up numbers to brag about; then use only 1:1000 targets. If want reality; use actual targets/samples like you are going to scan with your scanners; and you will get way lower numbers.</p>

<p>Here I really *want* to know the transfer function of my scanners; ie best case 1:1000 stuff and real world 1:10 and 1:15 contrast stuff of black and white and weird colors too. Knowing 1:1000 data is a best case number; 1:2 contrast stuff might be only 25 to 50 percent as good as resolution. It actually represents alot more valid to real world originals; but strangely few understand this. It is sort of like wanting real world gas mileage going to the store; work versus a super best case conditon that has high goosed numbers.</p>

[kelly_flanigan1]

<p>If one tests lenses or scanners with 1:1000 contrast references; one gets *way* higher "resolution numbers" than when lower contrast references are used.</p>

<p>Ponder: Two devices can measure the same resolution numbers with 1:1000 contrast targets/references; but one is *alot* better than another when a 1:5 contrast target/reference is used.</p>

<p>If the amateur thus uses only 1:1000 contrast targets with scanners or lenses; it is like only comparing best case highway mileage of Car A versus B.</p>

<p>One can have two cars with the same highway best case high numbers; but Car A might burn only 1/2 the gas in town versus Car B.</p>

<p>The simpleton assuming 1:1000 test of comparing scanners and lenses thus can be wonky with real world 1:5 contrast originals.</p>

<p>The better scanner of two with 1:1000 targets can be worse with 1:5 originals; ie the simple model of assuming 1:1000 fails to look the entire transfer function; you "cherry picked" one point on the transfer function curve and used it to state scanner A is better than C.</p>

<p>Thus is the reason so many folks get into a pickle; they run best case tests and like to assume stuff and their model is not the whole ball of wax; in fact it does not even do a test like a typical scan a client is going to bring in; it uses a goosed super high contrast 1:1000 target up in the stratosphere; not on the real earth.</p>

<p> </p>

[cliff_rames]

<p>Jay Bees - You could make a slanted-edge target with a razor blade or other straight edge and a slide mount. Then use IMATEST, or even an <a href="http://rsbweb.nih.gov/ij/plugins/se-mtf/index.html">ImageJ plugin</a> to derive the MTF.</p>