Ad
Ad
Ad
Pages: [1]   Bottom of Page
Print
Author Topic: Imatest recommendation  (Read 5967 times)
normkoren
Newbie
*
Offline Offline

Posts: 6


WWW
« on: May 16, 2006, 01:09:20 PM »
ReplyReply

Thank you very much for an extremely interesting comparison.

At the risk of appearing shamelessly self-serving (gotta take risks sometimes), I would like to recommend that you consider using Imatest for future tests. Of course, Imatest is not a substitute for real images, but it can provide superb quantitative comparisons between cameras and lenses. It can measure sharpness with far greater precision than you can estimate by observing images; you can measure exactly how much they differ.

Before I go too far, here is some background. Imatest is an affordable program for measuring digital image quality (especially sharpness, which is measured as MTF). It involves photographing and analyzing simple targets: slanted-edge for sharpness, step charts, the GretagMacbeth ColorChecker, and a few others. Here are some key links:

Home page:   http://www.imatest.com/
Image quality factors:  http://www.imatest.com/docs/iqf.html
Sharpness (explanation of MTF):  http://www.imatest.com/docs/sharpness.html
How to test lenses:  http://www.imatest.com/docs/lens_testing.html

A few points:

Imatest is particularly good for measuring lens sharpness (though it does so as a part of system sharpness; it can't measure lenses by themselves). All you need is to include a simple slanted edge in the photo (several if you want to see how performance changes at the edge). It's easy to find the optimum aperature and see how much sharpness is lost at non-optimum apertures. Lenses vary the most when they are wide-open. Most tend to be diffraction-limited at small apertures. The meaning of"Small aperture" (where sharpness starts dropping significantly) depends on the format. It's roughly f/11-f/16 for 35mm format; f/16-f/22 for larger formats.

Imatest produces excellent results regardless of image contrast-- you don't need to normalize image contrast, which is the correct thing to do for making visual comparisons. Normalization is built into Imatest.

Imatest shows exactly what sharpening does. It is valuable for comparing different raw converters.

Imatest can compare color accuracy using the ColorChecker images that will be on the disk. (It can also measure noise in the ColorChecker patches.) See http://www.imatest.com/docs/tour_colorcheck.html .

Imatest can measure dynamic range, but the measurement requires some care. Dynamic range for cameras in this class can be quite high: 11 to 13 f-stops, enough to push the limits of the Stouffer T4110 transmission step chart, which has a maximum density of 4.0 (about 13 f-stops).

Enough for now. There's lots of detail on the Imatest website. Yes, it's "pixel peeping," but it's done right and carried to the limit. Although I designed Imatest for serious photographers interested in testing their camera and lens systems (like most readers of this site), it has been widely adopted by the mobile imaging (i.e., camera phone) industry. Yes, I know, camera phone images are pretty bad (and they'll never be as good as large formats; size matters), but they will get better. About 600 million of the little things will be sold this year, making camera phones the fastest growing industry in history, at least for a year or two. It doesn't feel too bad to dominate image quality testing in that industry, especially since it happened by accident.
Logged

Gary Ferguson
Sr. Member
****
Offline Offline

Posts: 522


WWW
« Reply #1 on: May 18, 2006, 10:02:37 AM »
ReplyReply

Norman, I've always been impressed with your site and I've learnt a great deal from you, so I'd like to experiment with Imatest. However, I use Mac OSX.

The Imatest site says I'd need Virtual PC. I don't have much experience with Virtual PC, but when I have used it it seems to run at a snail's pace. Given that usually anything to do with images involves big files is it really a viable option?
Logged
normkoren
Newbie
*
Offline Offline

Posts: 6


WWW
« Reply #2 on: May 18, 2006, 10:49:21 AM »
ReplyReply

About a year ago I tried it with Virtual PC on an older Mac, and it seemed to work well. We plan a Mac version later in the summer. Of course the new Macs will be able to run Windows in native mode, and I hear they're excellent. The Matlab compiler has only recently supported the Mac, and I've been incredibly busy during that time. The Mac market is very important for individual photographers and for education, so it's high on our priority list. I do most of the programming myself. It's starting to look like I'll have to go the angel funding route so we can expand and complete all the projects on the (large) list.
Logged

Pages: [1]   Top of Page
Print
Jump to:  

Ad
Ad
Ad