Every infrared camera does it — that momentary freeze, the brief flash of grey, and then a sharp image returns. This is NUC: Non-Uniformity Correction. It’s one of the least-understood behaviors in thermal cameras, and one of the most consequential for product design. Here’s a complete breakdown.
Why NUC Is Necessary
No two pixels in a focal plane array are identical. Each pixel has slightly different:
- Responsivity (how much signal it produces per unit of incident radiation)
- Offset (its baseline output with no light)
- Noise characteristics
Without correction, an image from an uncalibrated array looks like a snowfield — each pixel at a slightly different brightness even when viewing a uniform scene. This is called fixed-pattern noise (FPN), and it’s unacceptable for useful imaging.
NUC algorithms measure and correct for these pixel-to-pixel variations using mathematical models applied per-pixel. The result is a flat-field corrected image where a uniform-temperature scene appears uniform in the output.
Two-Point NUC vs Multi-Point Calibration
Factory Calibration
When a module ships from IRmodules, it contains a factory-calibrated gain and offset table for every pixel. This calibration is performed at controlled temperature conditions using precision blackbody sources. The factory calibration table is valid at the reference temperature and degrades slowly as the detector ages.
In-Field NUC (Flat-Field Correction)
Because detector behavior drifts with ambient temperature and over time, periodic in-field NUC is needed. This is accomplished by placing a uniform-temperature reference in front of the detector — in uncooled LWIR modules, this is a mechanical shutter (flag) that swings briefly in front of the optical path.
With the shutter closed, every pixel should see the same temperature (the shutter temperature). Any pixel-to-pixel variation at this point is captured as a new offset correction. The shutter opens; the corrected image is applied.
This entire process takes 10–30 milliseconds and results in 1–3 lost video frames.
NUC Trigger Conditions
Uncooled LWIR modules typically trigger NUC under several conditions:
| Trigger Type | Default Condition | Configurable? |
|---|---|---|
| Temperature change | Every 2–5°C detector temp change | Yes |
| Time-based | Every 60–300 seconds | Yes |
| Power-on | Always | Fixed |
| Manual command | Via serial interface | Yes |
The temperature-change trigger is the most important: as the module’s operating temperature changes (aircraft altitude change, ambient temperature swing, sun exposure), the per-pixel characteristics drift and NUC must update to track.
Designing Around NUC in Your System
For most surveillance and situational awareness applications, automatic NUC is transparent — operators barely notice the brief freeze. But several applications require careful NUC management:
Video analytics and AI detection: A NUC event creates a missed frame, which triggers a false “target disappeared” event in many tracking algorithms. Solutions:
- Buffer video and fill NUC gaps with the last valid frame
- Suppress NUC trigger via serial command during active tracking events
- Use a short-exposure NUC algorithm that minimizes frame loss
Continuous recording: Most recording systems handle dropped frames gracefully, but timestamps in recorded data may show a gap. Metadata should mark NUC events explicitly.
Weapon sights and targeting: NUC at the moment of firing is unacceptable. Defense system designers typically implement manual NUC control: the system commands NUC when not in an engagement state and suppresses it during active target lock.
Scene-Based NUC
Some advanced imaging systems use scene-based NUC (SBNUC) — algorithms that derive correction coefficients from natural scene motion, without a mechanical shutter. SBNUC works by observing that pixels should have consistent relative responses as scene content moves across them. Advantages: no shutter mechanism (reduced complexity, no moving parts), no frame loss. Disadvantages: requires scene motion, takes longer to converge, and is computationally demanding.
IRmodules’ standard module series use shutter-based NUC. Scene-based NUC can be implemented in the host processor using module raw output — contact our engineering team for raw-data interface documentation.
Practical Recommendations
- For surveillance and ISR: accept automatic NUC, ensure your recording pipeline handles dropped frames gracefully
- For tracking and AI applications: implement NUC-aware frame management in your video pipeline
- For weapon sights: use manual NUC control via serial interface
- For all applications: verify NUC frequency in your thermal environment — a module exposed to direct sunlight in an outdoor enclosure may NUC more frequently than expected
