Free Resolution Testing with LCD Displays

[Andy Resnick]

I've done plenty of imaging performance testing on a variety of systems, from millimeter wave through UV. Until very recently, such testing required expensive and specialized equipment. Now, with digital imaging (at least in the visible), you can perform the same type of characterization on your camera for free.

All you need is an LCD display.

LCD displays work totally differently from old CRT displays- they are not raster-scanned, for example. So, you can take photos of the display without fear of flicker (which happens if the shutter speed is not a multiple or fraction of 1/30s).

Why do this? Lenses do not deliver constant performance over the full range of f-stop. Photographers often say things like "always use the lens stopped down at least 1 stop from maximum", and the reason is that almost all aberrations get worse with larger aperture (smaller f-stop). Maximum sharpness will be obtained for a limited range of f-stops: smaller f-stops show increased aberration, while larger f-stops show effects from diffraction through a pinhole (the tiny aperture). Knowing how your lens performs with aperture can help inform your choices when shooting.

Here's an example, using my 15mm lens. First, the full-frame: this is our LCD TV set on pause- the particular image was used because there is high-contrast features around the edges of the frame.

[PLAIN]http://img29.imageshack.us/img29/1141/presentation1sl.jpg

The key is to take an image with as much of the screen in the frame as possible- I am trying to have regular features at the resolution limit of my camera sensor. I've identified regions on the frame for closer examination: the center of the frame is almost always fully corrected, while the other points were selected to show performance at full image height, where aberrations are the worst.

Here's section 'A' at 100%,with left-to-right images taken at f/3.5 (full aperture), f/8, and f/16:

[PLAIN]http://img835.imageshack.us/img835/4705/37766796.jpg

Not much difference, as expected. The pixels are all rectangular (no distortion), there is little chromatic aberration (purple fringing), etc.

Here's section 'C', with the same apertures stacked from top to bottom:

[PLAIN]http://img137.imageshack.us/img137/4039/45029677.jpg

Now we can see a difference- at full aperture, the pixels are no longer so easily resolved, while at f/8 and up the pixels remain sharp. Again, there is no obvious distortion, astigmatism, or chromatic aberration- the likely culprit is uncorrected field curvature.

Here's section 'E':

[PLAIN]http://img854.imageshack.us/img854/9894/21068224.jpg

Again, lens performance is markedly improved at f/8 and f/16 over f/3.5. In fact, at f/3.5 the performance on this side of the image is worse than the other side- the lens performance is not quite symmetric, indicating an element may be slightly misaligned.

So why get a fast lens, if the performance becomes degraded at those low f-numbers? Lots of reasons. Personally, when I shoot at maximum aperture I'm less concerned with getting a tack-sharp image and instead more concerned with trying to capture a low-light level scene.

Another effect that's useful to understand in sampled imaging systems is 'aliasing'. This occurs because sampled imaging systems are no longer linear-shift-invariant, and the spatial frequencies present in the sensor can interfere with spatial frequencies from the object. Here's an example of aliasing, casued by re-sizing the full image:

[PLAIN]http://img835.imageshack.us/img835/400/aliasu.jpg

The phenomenon is identical to "Moire' patterns" and these greatly complicate quantitative measurements of imaging performance.

Hope this is useful...

 

[Ygggdrasil]

Nice post. Perhaps this could be updated and adapted into a nice Insights article.