Multiplicative Noise Model
Definition
The mathematical description of PRNU: each pixel's output is approximately the product of the true light signal and the pixel's gain factor (1 + K), where K is the PRNU component. This scaling with signal level distinguishes PRNU from additive shot noise.
- Applies to
- PRNU camera sensor fingerprint
- Formula
- Output ~= true signal x (1 + K)
- K represents
- PRNU gain factor component
- Distinguishes from
- Additive shot noise
Common questions
Why does the noise being multiplicative rather than additive matter for PRNU extraction?+
Because the noise scales with the signal level, analysts must normalise or flatten images by their brightness before comparing PRNU patterns, otherwise differences in scene lighting between images would masquerade as differences in the camera fingerprint.
What image regions give the weakest PRNU signal under this model?+
Very dark or saturated regions carry little true signal to multiply against, so the PRNU component in those areas is weak and unreliable, which is why examiners favour mid-tone, well-exposed regions when matching an image to a candidate camera.
Related terms
- DCNU (Dark Current Non-Uniformity)
- Additive noise generated by thermal activity in pixels even without light. Unlike PRNU, DCNU varies with temperature and camera gain settings, so...
- Fixed-Pattern Noise (FPN)
- Any repeatable, spatially correlated noise in an image sensor, as opposed to random shot noise. PRNU and DCNU are both forms of...
- Lens Vignetting
- The radial darkening toward the corners of an image caused by the optical path. It creates a low-frequency spatially correlated pattern in...
- PRNU (Photo Response Non-Uniformity)
- The unique pattern of pixel-level sensitivity variations in a camera sensor, used as a device fingerprint. A genuine camera-original image carries the...
- Sensor Pattern Noise (SPN)
- The composite fixed-pattern component in an image that includes PRNU plus lower-level contributions from optics and pixel defects. The PRNU component dominates...