How do you calculate the minimum detectable defect size?

First get resolution by dividing FOV by horizontal pixels, then multiply by the minimum pixels needed per defect. With 220 mm FOV, 2448 pixels and a 3-pixel rule: (220 / 2448) x 3 = 0.27 mm.

How many pixels do you need to detect a defect?

A common rule of thumb is 3 pixels for an easy high-contrast defect and 5 or more for subtle low-contrast flaws. The safer the application, the higher the pixel count you should require.

What is a good minimum detectable defect for vision inspection?

Good means comfortably smaller than the smallest defect the spec calls out. If the standard requires catching 0.5 mm flaws and your system resolves down to 0.27 mm, you have healthy margin.

Why use 3 pixels instead of 1 pixel per defect?

A single pixel of signal is indistinguishable from sensor noise and gets lost in interpolation. Requiring 3 or more pixels ensures a defect produces a consistent, segmentable blob the algorithm can trust.

How do I detect smaller defects?

Reduce the FOV with a longer lens or closer working distance, increase camera pixels, or improve contrast so you can lower the pixels-per-defect requirement. Halving the FOV to 110 mm drops the minimum defect to about 0.13 mm at the 3-pixel rule.

Minimum detectable defect vs measurement resolution?

Measurement resolution is one pixel at the part; minimum detectable defect is that times the pixels-per-defect rule. Detection always needs several pixels, so the detectable defect is larger than the raw resolution.

Machine Vision & Industrial Inspection AI calculator

Minimum Detectable Defect Size Calculator

The minimum detectable defect is the smallest physical flaw a vision system can reliably find, set by the pixel size at the part and how many pixels an algorithm needs to declare a defect. It is the make-or-break number in an inspection feasibility study: if your smallest required defect is below this threshold, no amount of clever code will save the application. Quality engineers and integrators compute it to decide whether a camera, lens and working distance combination can meet the inspection spec. This calculator multiplies the spatial resolution by the minimum pixels-per-defect rule to give the smallest catchable defect in mm.

What this calculator does

Calculate the minimum defect size a machine vision camera can reliably detect by multiplying the inspection resolution (mm per pixel) by the minimum number of pixels required to detect and classify a defect.
Use it when writing or validating a machine vision specification to confirm that the camera and lens combination can detect the smallest defect in the inspection requirement.
It computes the smallest reliably detectable defect size in mm by multiplying spatial resolution (FOV / pixels) by the minimum pixels needed to register a defect.

Formula used

Inspection resolution = FOV width / horizontal pixel count
Minimum detectable defect size = inspection resolution x minimum pixels required
Example: (220 mm / 2448 pixels) x 3 = 0.27 mm minimum defect size

Inputs explained

Horizontal field of view (FOV width):
Camera horizontal pixel count:
Minimum pixels required to detect a defect:

How to use the result

Use it during application feasibility to confirm the optics can resolve the smallest defect the customer or standard requires.
The pixels-per-defect rule is empirical and depends on contrast, lighting and algorithm; low-contrast scratches may need more pixels than high-contrast holes, so treat the result as a best case.

Common questions

How do you calculate the minimum detectable defect size? First get resolution by dividing FOV by horizontal pixels, then multiply by the minimum pixels needed per defect. With 220 mm FOV, 2448 pixels and a 3-pixel rule: (220 / 2448) x 3 = 0.27 mm.
How many pixels do you need to detect a defect? A common rule of thumb is 3 pixels for an easy high-contrast defect and 5 or more for subtle low-contrast flaws. The safer the application, the higher the pixel count you should require.
What is a good minimum detectable defect for vision inspection? Good means comfortably smaller than the smallest defect the spec calls out. If the standard requires catching 0.5 mm flaws and your system resolves down to 0.27 mm, you have healthy margin.
Why use 3 pixels instead of 1 pixel per defect? A single pixel of signal is indistinguishable from sensor noise and gets lost in interpolation. Requiring 3 or more pixels ensures a defect produces a consistent, segmentable blob the algorithm can trust.
How do I detect smaller defects? Reduce the FOV with a longer lens or closer working distance, increase camera pixels, or improve contrast so you can lower the pixels-per-defect requirement. Halving the FOV to 110 mm drops the minimum defect to about 0.13 mm at the 3-pixel rule.

Last reviewed 2026-05-12.