MECH5720 Chap.2 Image Formation and Sensor Physics
Image Formation and Sensor Physics
Image formation connects ray geometry to a sampled, noisy electrical measurement. The thin-lens equation predicts ideal conjugate distances and magnification, while aperture controls both collected light and the cone of accepted rays. Pixels integrate photons over area and exposure time, then electronics add offset, read noise, gain and saturation limits.
The useful model therefore progresses from rays and irradiance to charge, quantisation and dynamic range without confusing an ideal lens calculation with a complete camera calibration. A ray diagram provides the first geometric check. Rays from an object point are refracted so that they converge near an image point, with the optical axis and principal-plane convention defining the signed distances.
The thin-lens equation supplies a first estimate of sensor location, and magnification predicts image orientation and scale. A real compound lens has thickness, aberration and manufacturing tolerances, so the calculation should guide a focus sweep rather than replace it. Aperture changes the accepted cone of rays.
Opening it admits more light and can shorten an exposure, but also changes depth of field and the influence of aberrations. Closing it can improve some geometric errors until diffraction spreads fine detail. F-number relates effective aperture diameter to focal length and is more informative than diameter alone when two lenses are compared. The sensor converts irradiance over area and time into charge.
Photon arrival has shot-noise variation; readout adds electronic uncertainty; dark signal and fixed response create structured terms; quantisation maps an analogue value to discrete levels. Longer exposure generally raises photon count and shot-noise signal-to-noise ratio, yet motion and saturation impose limits. Gain changes the recorded scale without creating photons and can also alter the effective read-noise contribution.
Sampling connects the optical image to a pixel grid. Pixel pitch, magnification and target distance determine the scene spacing represented by one sample. If optical content exceeds the sampling limit, aliases appear as false lower frequencies. An optical low-pass effect can prevent folding, whereas interpolation applied after sampling cannot identify which high frequency produced the alias.
Calibration should include dark frames, a uniform-field response where needed and a known geometric target. Repeat the measurement at relevant exposure and temperature because offsets and noise behaviour may change. Report the operating range between noise floor and saturation, not just nominal bit depth.
A defensible image-formation explanation ends by identifying which stage currently limits the task: geometry, optical transfer, sampling, photon statistics, electronics or model mismatch.
What this chapter covers
- 01
Pinhole projection
- 02
Thin-lens geometry
- 03
Magnification and field of view
- 04
Photon and read noise
- 05
Pixel sampling
- 06
Dynamic range and saturation
Worked example · free
Compute image distance and scale
- +1Use 1/di = 1/35 - 1/700 in reciprocal millimetres.
- +1The reciprocal is 19/700, giving di = 36.84 mm.
- +1Compute m = -di/do = -36.84/700 = -0.0526.
- +1Interpret the sign as inversion and the magnitude as reduction.
Key terms
- Focal length
- A lens parameter relating object and image distances in the thin-lens model.
- Magnification
- The signed ratio of image size to object size.
- Shot noise
- Photon-count fluctuation whose variance grows with expected count.
- Read noise
- Electronic uncertainty introduced during sensor readout.
- Dynamic range
- The usable span between the noise floor and saturation.
Image Formation and Sensor Physics FAQ
What does the thin-lens equation omit?
It omits lens thickness, principal-plane detail, aberrations, diffraction, field dependence and assembly tolerances. It is valuable for first-order geometry, while physical calibration locates best focus and characterises departures in the real camera.
Why does collecting more light not remove every noise problem?
More photons improve shot-noise signal-to-noise ratio until saturation or motion becomes limiting, but read noise, dark offset, fixed-pattern response and model errors remain. Exposure decisions must consider the full measurement chain.
How do pixel size and optics interact?
Pixels sample the optical image, so useful resolution depends on both optical transfer and sampling density. Smaller pixels can sample more finely but collect fewer photons per pixel for the same irradiance and exposure.
Exam move
Practise thin-lens calculations with millimetres, metres and changed sign conventions until unit conversion is automatic. Then attach each ideal result to one physical qualification. Build a table that separates photon noise, read noise, quantisation and saturation by cause, scaling behaviour and acquisition remedy. Sketch three rays before substituting numbers so the sign and inversion have a geometric check.
For sensor questions, trace photons to charge, gain and digital value instead of treating noise as one term. Compare a longer exposure, wider aperture and higher gain by naming what each changes physically. Finish with a calibration plan containing a dark frame, a known target, a focus sweep and a saturation check under the intended operating conditions.
Explain which observation would show that nominal bit depth overstates the usable dynamic range.