The most common cause of failed CCTV-based investigations is not poor camera placement or inadequate coverage — it is black-and-white footage captured in conditions where colour was available but the camera's sensor couldn't use it. When a robbery occurs at 11pm in a car park with ambient streetlight, the suspect's red jacket, blue vehicle, and distinct yellow bag are all recorded as indistinct grey shapes. Witnesses describe vivid colours; the CCTV shows nothing. ColorVu and equivalent full-colour low-light cameras close this gap permanently.
Three converging technologies make full colour at near-zero lux possible: F/1.0 or F/1.2 maximum aperture optics that gather 4× more light than standard CCTV lenses; Sony STARVIS 2 Back-Side Illuminated (BSI) CMOS sensors with significantly improved quantum efficiency; and precisely tuned supplemental white-light LED illuminators that add visible spectrum illumination without the glare or colour cast of poorly designed white-light cameras.
Low-Light Technology Comparison
| Technology | Min. Illumination | Colour Performance | IR-Cut Filter | White-Light Range | Typical Application |
|---|---|---|---|---|---|
| Standard WDR camera | 0.01 lux colour | Colour day / mono night | Switches to IR mode | N/A | Well-lit interiors |
| Sony STARVIS CMOS | 0.005 lux colour | Low-light colour extension | Reduced switch threshold | N/A | Low-light coverage |
| Hikvision ColorVu 2.0 | 0.0005 lux colour | Full colour 24/7 | No IR mode — always colour | 20–60m | Car parks, perimeter |
| Axis Lightfinder 2.0 | 0.003 lux colour | Full colour low-light | Intelligent mode selection | N/A (sensor only) | Indoor, ambient lit |
| Dahua Full-color | 0.001 lux colour | Full colour with LED | Warm white LED supplement | 30–50m | Entrances, retail |
| AI-enhanced denoising | 0.0001 lux (AI) | AI-recovered colour detail | Not applicable | N/A | Research / premium |
The Sensor Technology Behind Full-Colour Low-Light
- Sony STARVIS 2 BSI: Back-Side Illumination (BSI) places photodiodes on the front face of the silicon, behind the microlens array — removing the wiring layer that blocks light in front-illuminated (FSI) sensors. BSI improves quantum efficiency (photons captured per photon received) by 50–80%, directly translating to better low-light sensitivity. STARVIS 2 adds enhanced NIR response and reduced read noise compared to the original STARVIS.
- Large-aperture optics: Aperture area is proportional to the square of the f-number ratio. An F/1.0 lens gathers 4× more light than an F/2.0 lens (2² ratio). F/1.0 lenses for CCTV cameras are precision-ground elements with tight manufacturing tolerances — significantly more expensive than standard CCTV lenses but essential for sensor-only (non-white-light) colour low-light performance.
- Pixel binning: In extreme low-light, camera ISPs combine adjacent pixel values (2×2 or 4×4 binning) — trading resolution for sensitivity. A 4K camera in 4×4 binning mode produces 1MP output but with 16× the light-gathering of an unbinned pixel. Modern cameras perform adaptive binning — full resolution in adequate light, automatic binning as light falls below threshold.
- White-light LED design: Poor white-light cameras create harsh flat illumination and glare. Quality ColorVu implementations use multiple lower-power LEDs distributed around the lens — creating even, directional illumination that preserves scene depth and reduces specular reflection from wet surfaces and glass.
AI Computational Photography: Identification-Quality Colour at 0.0001 Lux
By 2028, AI-based computational photography will enable CCTV cameras to produce identification-quality colour footage at illumination levels approaching 0.0001 lux — through multi-frame HDR stacking (combining 30+ frames to extract colour signal from noise), AI denoising models trained on millions of low-light image pairs, and neural super-resolution that recovers facial detail from sub-threshold sensor data. The result is a camera that records usable forensic colour evidence in environments that a human eye perceives as effectively pitch-black. The boundary between night and day disappears entirely from a forensic evidence perspective.