A thermal imaging module or thermal camera core only wins in the field when its palette makes targets “pop” without eye strain. In any thermal imaging camera module, the palette is the last mile of perception: it translates 14-bit radiance into an 8-bit picture humans can parse fast, on OLEDs, phones, or VMS streams. This guide turns palette choice into numbers you can spec, test, and defend.
Executive Summary
Good palettes are task-driven, not fashionable. White-hot and black-hot cover 80–90% of use because they are monotonic and low-cognitive-load. Iron/amber variants help when you need to separate warm foreground from warm background. “Rainbow” looks dramatic but hides edges and exaggerates noise—avoid for shipping defaults. Lock AGC, palette, and gamma together; then validate legibility under bright sun and after NUC events. Publish an acceptance card with contrast targets, just-noticeable steps, and eye-fatigue minutes so buyers can reproduce your claims.
Use Cases & Buyer Scenarios
Scenario 1 — Handheld observation and spotting
Hunters, rangers, and technicians need fast detection at dawn/dusk and noon. Defaults: white-hot outdoors; black-hot at night to reduce glare; iron/amber when background runs hot (summer rocks/roofs). Align HUD habits with the families under Thermal Optics so users switch seamlessly between Thermal Monoculars and Thermal Rifle Scopes.
Scenario 2 — Security VMS and industrial monitoring
Operators stare at walls of video; eye strain matters more than drama. Ship white-hot, black-hot, and a muted ironbow with a fixed AGC ladder for each stream profile. ONVIF/RTSP devices should expose palette as metadata; sample profiles and docs live on Support.
Scenario 3 — UAV search & mapping
Moving backgrounds and sun-glint punish aggressive AGC and flashy palettes. Use white-hot with clipped local AGC and a thin outline overlay; switch to amber/iron for hot terrain. For fusion projects with range data, see Thermal + LRF Fusion & Ballistics.
Spec & Selection Guide
What a “palette” really is
A palette is a 256-entry look-up table (LUT) that maps a normalised thermal value v∈[0,1] to RGB. The scene → value path is: sensor (14-bit) → non-uniformity correction (NUC) → AGC (contrast stretch or local) → gamma → LUT → display/compressor. Your product “look” is the bundle of AGC + LUT + gamma + text/reticle rendering.
Key definitions
Monotonic palette. As temperature rises, pixel luminance rises (white-hot) or falls (black-hot). Monotonic = predictable edges. Perceptual uniformity. Equal value steps look like equal brightness steps. Palettes like viridis are designed for this. AGC (automatic gain control). Maps scene histogram to the 0–255 display range. Global AGC is stable; local AGC increases local contrast but can “pump” and create false edges. JND (just-noticeable difference). One perceptual step. A good palette/AGC combo provides ≥60–80 JNDs across the important portion of the scene histogram.
| Palette | Strength | Watch-outs | Recommended use | Encoder efficiency |
|---|---|---|---|---|
| White-hot | Fast edges, low cognitive load | Glare at noon if UI too bright | Default for most outdoor scenes | High (simple gradients compress well) |
| Black-hot | Lower eye strain at night | Warm sky can reduce separation | Night, security operators | High |
| Iron/amber | Separates warm subject from warm background | Over-saturation hides edges | Urban roofs, summer rocks, HVAC | Good |
| Sepia | Calm, low-glare UI look | Weaker small-contrast cues | Extended monitoring shifts | Good |
| Rainbow | Intuitive for demos | Non-uniform; false edges; “banding” | Avoid as default; lab only | Fair (color flicker ↑ bitrate) |
Rules of thumb you can put in a PRD
- Ship three defaults only: white-hot, black-hot, and iron/amber. Keep others in “advanced.”
- Pair each palette with a named AGC preset (Global-Stable, Local-Detail, Clip-Hot) and a gamma (e.g., 1.8–2.0).
- Set UI contrast targets: digits/reticle ≥4.5:1 in ≥100 klx sun; ≥10:1 at night on OLED.
- Never let palette switching change compression ladder or latency unexpectedly.
Integration & Engineering Notes
Electrical & Interfaces
Expose palette/AGC via UART or CAN with idempotent commands (GET/SET_PALETTE, GET/SET_AGC, GET_GAMMA). For streaming, surface palette choice in ONVIF metadata and keep uniform colorimetry across ladders. Sample message maps are in Downloads.
Optics & Mechanics
Palettes cannot fix poor readability caused by windows and eyepieces. Use hard AR coatings and hydrophobic tops to keep contrast in rain and condensation. The UI discipline should mirror what we ship across Products and Modules.
Firmware/ISP/Tuning
- Global-Stable AGC. Percentile-based stretch (e.g., map 2–98% to 0–255) with mild slope limiting. Best for moving cameras.
- Local-Detail AGC. CLAHE-style local contrast for stationary cameras; cap tile gain to prevent halos and pumping.
- Clip-Hot. When a few pixels saturate (e.g., exhaust, sun reflections), clip <1% of the brightest bins and spread mid-tones.
- Gamma. 1.8–2.0 gives pleasing mid-tones; expose “Gamma: Low/Std/High” rather than numbers to end users.
- NUC events. During shutter/FFC, dim or freeze OSD and reduce encoder QP sensitivity to avoid bitrate spikes labeled as “glitches.”
- Palette persistence. Save per-mode (handheld/vms) and survive power cycles; log changes in a circular buffer.
Testing & Validation (bench → field)
Acceptance gates (illustrative). Readability contrast ≥4.5:1 at ≥100 klx; ≥60 JND steps across 5–95% histogram; detection time ≤2.0 s for a 2 mrad target at 100 m; eye-fatigue score ≤3 on a 1–7 Likert after 20 min viewing; palette swap latency <100 ms end-to-end; encoder bitrate variance <25% across palettes at fixed scene.
Bright-sun readability. Use a sun-lamp box; measure on-camera contrast of digits/reticles in each palette with a photometer. JND staircase. Render a greyscale ramp and a uniform-temperature panel; count visually separable steps for each palette/AGC. Operator task test. Timed “find the animal” or “find the leak” tasks using controlled clips; log time-to-first-fix and errors. NUC resilience. Trigger a shutter; confirm encoder doesn’t spike >2× nominal and UI stays readable. VMS interop. RTSP/ONVIF playback in Milestone/Exacq and mobile apps; palette metadata visible; no color range mismatch.
Compliance, Export & Certifications
Palettes themselves are not regulated, but your streaming hardware and EMC are. Keep CE/FCC/UKCA marks, RoHS material declarations, and (if combined with LRF) ensure Class-1 emissions per IEC 60825-1 remain invariant with UI modes. Public certs and statements live on Certificates and terms on Warranty.
Business Model, MOQ & Lead Time (OEM/ODM)
We ship palette/AGC presets tuned for handheld and VMS use, plus a small “operator study” kit (clips, tasks, scoring sheet) you can reuse with channels. Typical MOQs: 100–300 pcs for catalog optics; 500–1,000 pcs for custom encoders or unique palettes. EVT in 4–6 weeks; custom glass/coatings add 6–10 weeks. Explore options under Module Integration for OEMs and see families at Thermal camera module.
| Deliverable | What it is | Channel impact |
|---|---|---|
| Palette/AGC acceptance card | Readability, JND, latency, bitrate variance | Faster retailer QA, fewer disputes |
| Operator study kit | Clips, tasks, scoring template | Evidence for tenders and demos |
| ONVIF sample XML + RTSP traces | Profile S/T configs with palette metadata | Plug-and-play with VMS |
Pitfalls, Benchmarks & QA
- Shipping “rainbow” as default. It is non-uniform; it creates false edges and eye fatigue.
- Letting AGC pump. Local contrast without gain capping flickers during motion; cap CLAHE tile gain and add temporal smoothing.
- Palette changes that break recording. Range mismatches (studio vs full) or sudden GOP resets make VMS files ugly; keep colorimetry and keyframe cadence stable.
- No sun test. If digits fail at ≥100 klx, the palette debate is moot; fix luminance and contrast first (HUD auto-dim with floor).
- Ignoring eye-fatigue. Ten minutes might look fine; 30 minutes reveals strain—test long sessions.
FAQs
Q: Which single default should I ship?
White-hot with Global-Stable AGC and gamma 1.9. It wins most scenes and compresses well.
Q: When does black-hot beat white-hot?
At night and in security rooms—dark backgrounds and UI ergonomics reduce glare and strain.
Q: Is ironbow better for leak detection?
Often yes, because warm backgrounds compress dynamic range; amber hues separate subject edges without harsh saturation.
Q: Should I expose a “rainbow” option?
Only in “advanced,” with a warning. If you need color, use perceptually uniform maps (viridis/magma) with care.
Q: How do I keep VMS streams consistent?
Lock palette, AGC, gamma, and encoder ladder per profile; stamp palette in ONVIF metadata; test with two major VMS suites.
Decision Flow — from scene to shipping default
Start ├─ Task? detection / recognition / monitoring ├─ Background temp? cool / warm / mixed ├─ Motion? moving camera / static camera ├─ Pick AGC: │ • moving → Global-Stable (2–98% stretch, slope-limited) │ • static → Local-Detail (CLAHE, capped gain) ├─ Pick palette: │ • default → White-hot │ • night / operator room → Black-hot │ • warm background → Iron/amber ├─ Set gamma 1.8–2.0; HUD contrast ≥4.5:1 @ ≥100 klx ├─ Validate JND ≥60; palette swap < 100 ms; NUC resilience OK └─ Publish acceptance card + sample clips → Downloads → train channel
Call-to-Action (CTA)
Need palettes your users will actually keep? We’ll tune AGC and color ladders for your scenes, validate readability in sun and on VMS, and ship an acceptance card plus demo clips. Start a review via Contact, see hardware families under Thermal camera module, and explore integration services at Module Integration for OEMs.
Sources
- Choosing Colormaps. Guidance on perceptually uniform maps and pitfalls of rainbow. (Matplotlib)
- MIL-STD-1472H — Human Engineering. Display legibility and operator workload principles. (U.S. DoD QuickSearch)
- ISO 9241-303 — Electronic visual displays. Luminance/contrast requirements for readability. (ISO)
- Teledyne FLIR — AGC & Palettes Application Notes. Practical palette/AGC behavior in uncooled cores. (Teledyne FLIR)
- Edmund Optics — Image Contrast & Visual Perception. Human-factors context for edge visibility. (Edmund Optics)
