Monday, December 6, 2010

Angular resolution

Got a 70-210mm telephoto zoom for my camera and was wondering if that's enough. Did some research and calculations for angular resolving power, here are my results. They're expressed in milliradians (.001 radian) and sorted in descending order.

DeviceAng. res. (mrad)
iPhone 4 camera sensor (3.85mm lens, 1.4μm sensor pitch)0.72
iPhone 4 screen (326 dpi), 10 inches away0.62
22" 1680x1050 monitor, 3 ft away0.62
42" 720p TV, 10 ft away0.52
iPhone 4 camera lens (3.85mm f/2.8 aperture diffraction limit)0.51
42" 1080p TV, 10 ft away0.32
20/20 vision0.29
Human fovea (2 μm cone pitch, 17mm focal length)0.24
20/16 vision (typical healthy vision)0.23
210mm lens, 35mm film, 1.6 MP scan0.22
210mm lens, 75% contrast on Velvia 500.16
210mm lens, 25% contrast on Velvia 500.07
210mm lens, 35mm film, 4000 dpi scan0.06
210mm f/5.6 aperture diffraction limit0.04

I've tried to make this a fair comparison:
  • Diffraction limits (for a given aperture) are evaluated using the Rayleigh criterion at 580 nm. It's the minimum distance that two points of lights can still be resolved as separate points after diffraction. 
  • Pixel resolution (screens, sensors, scans, retina) are evaluated as 2*pitch/distance. Like the diffraction limit and visual acuity tests, it represents the closest that two lines or points can be distinguished from each other. In the frequency domain, this is the Nyquist frequency.
  • 20/20 vision is defined as being able to resolve lines with a spacing of 1 arcminute (approx 1/16" at 20 ft). Contrary to popular belief, it is a measure of resolving power, not myopia (though uncorrected myopia will reduce your resolving power at long distances).
  • Film resolving power comes from the modulation transfer function (MTF) in the datasheet. It measures contrast as a function of spatial frequency.
The one metric that stands out as unfair is the diffraction limit. For these optical systems, other lens aberrations are likely to dominate the resolving power. For normal lenses on 35mm, people only start to worry about diffraction around f/22, and that's partly because the effect of lens aberrations is mitigated at smaller apertures. It might come into play sooner for the smaller format/shorter lens of the iPhone camera.

Conclusions

  • If you can see it (naked eye), then it'll show up on a 1.6 MP image from a 210mm lens. Whether or not you get additional detail at higher resolution depends on the lens's optical quality.
  • There's no reason to want resolution greater than 1080p unless your TV is huge or you sit really close.
  • Apple's claims regarding the iPhone "retina" display are based on pitch/distance, not 2*pitch/distance. I believe this is flawed. I think the metric should be whether it can produce a set of lines that is just barely distinguishable from a solid gray block. Then the human eye would provide any anti-aliasing you might need.
  • The signal processing in your eye is quite good!

No comments:

Post a Comment