QQ: 859818709

Pixel Size and Resolution Calculator

Pixel Size and Resolution Calculator

📷 Pixel Size ↔ Resolution (lp/mm) Calculator

Pixel Size ↔ Resolution (lp/mm) Calculator

Convert bidirectionally between camera pixel size (pixel pitch) and spatial resolution (line pairs/mm, lp/mm), and evaluate whether the camera sensor resolution matches the optical system MTF cutoff frequency, quickly identifying the system bottleneck. Suitable for machine vision, laser inspection imaging and camera selection.

📷
Nyquist resolution
fN = 500 / p
At least 2 pixels are needed per line pair
Pixels per millimetre
Npix = 1000 / p
p is in μm
Optical MTF cutoff frequency (matching criterion)
fc = 1 / (λmm · F/#)
fN ≥ fcOptical bottleneck | fN < fcSensor bottleneck

📷 Pixel Size ↔ Resolution Calculator

Three modes: calculate resolution from pixel size, derive the required pixel size from a target resolution, and evaluate sensor-to-optics MTF matching




μm

Centre-to-centre spacing of adjacent pixels; see ‘pixel size / pixel pitch’ in the camera datasheet

Effective resolution reduction factor, typically 0.7~0.9; the default 1.0 means no reduction


📊 Calculation results

🎯 Nyquist resolution fN
—
lp/mm (line pairs/mm)
🔢 Pixels per millimetre
—
pixels/mm (dots/mm)
✏️ Minimum resolvable single-line width
—
μm (line-pair period = 2p)
📉 Effective resolution (×Kell)
—
lp/mm (equal to f when Kell=1.0N)

Formula details

From pixel pitch to Nyquist resolution, then to the optical diffraction MTF cutoff and the system matching criterion

① Definition of pixel size (pixel pitch)
Npix = 1000 / p
p: pixel size (μm), i.e. the centre-to-centre spacing of adjacent pixels (pixel pitch). The number of pixels per millimetre (dots/mm) = 1000/p. For example, when p=5 μm there are 200 pixels per millimetre.
② Nyquist sampling (1 line pair = 2 pixels)
fN = 1000 / (2·p) = 500 / p
A line pair consists of one bright line and one dark line; to resolve both, at least 2 pixels are needed to sample them separately. Therefore the Nyquist limit frequency (lp/mm) equals half the number of pixels per millimetre.
③ Meaning of lp/mm (line pairs/mm)
1 lp/mm ⇒ line-pair period = 1 mm
100 lp/mm means there are 100 bright–dark line pairs per millimetre, with a line-pair spatial period of 1/100 mm = 10 μm (bright and dark lines each about 5 μm wide). The larger the lp/mm value, the finer the detail that can be resolved.
④ Minimum single-line width
wmin≈ p, line-pair period = 2p
At the Nyquist limit, a single bright or dark line corresponds to about one pixel width p, while a complete line pair occupies 2 pixels with a period of 2p. This is the theoretical limit of ideal sampling.
⑤ Deriving the required pixel size
p = 500 / fN
Given the target resolution f required by the inspection taskN(lp/mm), the upper limit on the camera pixel size can be deduced. For example, if 100 lp/mm is required, the pixel size must be ≤ 5 μm.
⑥ Optical MTF cutoff frequency and matching criterion
fc = 1 / (λmm · F/#)
The MTF cutoff spatial frequency of an optical system at the diffraction limit, λmm = λμm/1000。fN ≥ fc: the pixels are sufficient to sample the optical limit, so the system is optics-limited (optical bottleneck);fN < fc: the pixels are too coarse and under-sampling occurs (sensor bottleneck).
⑦ Kell factor (effective resolution reduction)
feff ≈ fN × Kell
Real cameras are affected by pixel aperture shape, demosaicing algorithms and phase, so the effective resolution is typically only 0.7~0.9 times the Nyquist limit; this reduction ratio is the Kell factor. For rigorous selection, a value of 0.7~0.8 is recommended to leave a margin.

Parameter details

Three key quantities: pixel size, Nyquist frequency and optical MTF cutoff frequency

🔲

Pixel size p

  • Unit: μm; listed in datasheets as ‘pixel size / pixel pitch’
  • Common industrial cameras: 1.4 ~ 7.4 μm
  • Smaller p → more pixels per millimetre → fNhigher
  • but small pixels collect less light, and the signal-to-noise ratio and dynamic range usually decrease
🎯

Nyquist frequency fN

  • Unit: lp/mm (line pairs/mm)
  • fN= 500/p, the theoretical upper limit of discrete sampling
  • Note: pixels/mm (dots/mm) is twice the lp/mm value; do not confuse them
  • Actual effective resolution ≈ fN × Kell(0.7~0.9)
🔭

Optical MTF cutoff frequency fc

  • Unit: lp/mm, fc = 1/(λmm·F/#)
  • The shorter the wavelength and the smaller the f-number, the higher the optical cutoff frequency
  • At λ=0.55 μm and F/5.6, fc ≈ 324.7 lp/mm
  • compared with fNcomparing them tells you which end is the system bottleneck

📋 Quick-reference table of common pixel sizes

Comparison of pixel size, pixels per millimetre, Nyquist resolution and minimum single-line width

Pixel size p (μm) Pixels per millimetre (pixels/mm) Nyquist fN(lp/mm) Minimum single-line width (μm) Typical sensor reference
1.4 714.3 357.1 1.4 High-resolution small-pixel inspection camera
2.0 500.0 250.0 2.0 High-speed / high pixel-density camera
2.4 416.7 208.3 2.4 Common 1/1.8″~1″ industrial cameras
3.45 289.9 144.9 3.45 Classic pixels of the Sony IMX series
4.8 208.3 104.2 4.8 Large-format high-SNR camera
5.0 200.0 100.0 5.0 General-purpose machine vision camera
5.5 181.8 90.9 5.5 Low-light / high-speed camera
7.4 135.1 67.6 7.4 Large-pixel high-sensitivity camera

⚠️ Common pitfalls and notes

Six most common pitfalls in pixel-size/resolution conversion

Mistaking pixels (dots/mm) for line pairs lp/mm

The two differ by a factor of exactly 2. Pixels per millimetre = 1000/p, while one line pair requires 2 pixels, so fN= 500/p. At p=5 μm this is 200 pixels/mm, but only 100 lp/mm.

Do smaller pixels always give better imaging?

Not necessarily. As pixels get smaller, fNgets higher, but the actual system resolution is also constrained by the optical MTF cutoff frequency; if the lens resolution is insufficient, even the smallest pixels only produce ‘oversampling blur’, and small pixels also have poorer signal-to-noise ratio.

Ignoring the effect of wavelength and f-number on the optical cutoff

fc= 1/(λ·F/#). The longer the wavelength and the smaller the aperture (larger f-number), the lower the optical cutoff frequency. The optical resolution of blue light at F/2.8 and infrared at F/8 can differ by several times.

Confusing ‘line’ with ‘line pair’

In the TV-lines/TVL context, a ‘line’ sometimes means a single line, whereas lp/mm explicitly means a bright–dark pair of lines. The minimum single-line width ≈ p and the line-pair period = 2p; when quoting specifications, always check the unit definition.

Looking only at the total pixel count, not the individual pixel size

At the same 12 megapixels, the pixel size of a 1/1.7″ small format and a 1.1″ large format differs greatly, directly determining the resolution per millimetre and low-light performance. Spatial resolution depends on p, not on the total pixel count.

Treating the Nyquist limit as the actually achievable resolution

fNis the theoretical upper limit for ideal sampling; the actual effective resolution is about fN×Kell (0.7~0.9), with further degradation from lens MTF, alignment and motion blur; a margin should be left when selecting.

💡 Common applications

🤖
Machine vision camera selection
Based on the minimum defect size and field of view required by the inspection task, first determine the needed lp/mm (derive p using Mode B), then select a camera whose pixel size is no larger than that value, while also considering frame rate, signal-to-noise ratio and cost.
🧩
Optics / sensor matching assessment
Using Mode C, enter the pixel size, wavelength and f-number together and compare fNwith fc: a ratio far below 1 means the camera is holding back the lens (use smaller pixels); far above 1 means the lens is the bottleneck (open the aperture or use a higher-resolution lens).
🔬
Microscopy imaging resolution chain analysis
In microscopy/laser inspection imaging, the system resolution is determined jointly by the objective diffraction limit, relay optics, camera pixels and the Kell factor. Calculating stage by stage can locate the weakest link in the resolution chain, avoiding blind upgrades of a single component.

📋 Frequently asked questions

Q: Why does one line pair need 2 pixels?
A line pair contains one bright and one dark line. If sampled with only 1 pixel, the bright and dark are averaged into grey and cannot be distinguished; at least 1 pixel must fall on the bright line and 1 on the dark line to reconstruct the periodic fringes — this is the Nyquist sampling theorem.
Q: How do you convert between lp/mm and pixels/mm (PPI-type metrics)?
pixels/mm = 1000/p; lp/mm = pixels/mm ÷ 2 = 500/p. That is, the pixels/mm value is always twice the lp/mm value. When converting, combine with the Kell factor (0.7~0.9) to obtain the actual effective resolution.
Q:fNfar greater than fc— is it wasteful?
Moderate oversampling (fN/fcabout 1.5~2×) is not wasteful: it reduces aliasing, improves sub-pixel measurement accuracy and leaves margin for the Kell reduction; but beyond about 3× the returns diminish, and at that point the bottleneck is on the optics side — upgrading the lens is more effective than further shrinking the pixels.
Q: Why must the actual resolution be multiplied by the Kell factor?
Pixels have a finite aperture area, Bayer arrays require demosaicing, and there is a phase difference between the fringes and the pixel grid; all of these prevent details exactly at the Nyquist frequency from being reconstructed reliably. Empirically, the effective resolution is about 0.7~0.9 of the Nyquist frequency.

中文EN