1. Definition of Flicker

Flicker refers to the phenomenon of periodic brightness fluctuations, banding, or local brightness instability in images captured by a camera. This phenomenon is essentially caused by a mismatch in temporal sampling. LED lights typically emit light in pulses at frequencies of hundreds of times per second, and their apparent brightness is controlled by adjusting the duty cycle of the pulses (i.e., the proportion of the “on” time within a single cycle). Since the human eye acts as a natural temporal low-pass filter, these high-frequency pulses are visually “smoothed” into a continuous and stable light effect, making the LED light appear constantly on to the human eye. However, a camera capturing the light source may expose at the same temporal scale as these fluctuations, making it more susceptible to capturing the dynamic changes of the light source itself.

Frame N	Frame N+1

As shown in the figure above, in Frame N, the camera's exposure time overlaps with the pulse time of the PWM-driven LED traffic light, and the camera captures the red light; in Frame N+1, the two do not overlap, and the red light is not captured. Consequently, the traffic light appears to turn on and off depending on whether the camera's exposure time overlaps with the LED light pulses in consecutive frames, resulting in the flicker phenomenon.

2. Types of Flicker
1. Illuminant Flicker
Flicker of directly imaged light sources within the camera's field of view (FOV), such as headlights, traffic lights, or road signs, is referred to as “illuminant flicker”.

Figure 2 Illuminant flicker

When headlights are activated, the duty cycle and frequency of the LEDs are adjusted to dim the lights. As shown in the figure above, the LED on the left appears off while the one on the right appears on. This is because the headlights of most vehicles are not synchronized in frequency, duty cycle, or phase, causing them to flicker at different rates or phases.

2. Reflectance Flicker
Flicker caused by the illumination of flickering light sources outside the camera's field of view (FOV), such as the flicker presented by surface reflections when lights illuminate the road, is referred to as “reflectance flicker”.

Figure 3 Reflectance flicker

The figure above shows banding artifacts captured by a rolling shutter camera when imaging a uniform target illuminated by a PWM flickering light source, taken at various frequencies between 100 Hz and 1000 Hz. Unlike 50/60 Hz AC banding effects, the number and height of the bands vary depending on the LED's frequency and duty cycle, as well as the camera's frame rate.

3. Test Equipment and Setup
“Illuminant Flicker”:
The PWM-driven LED light source should be placed within the field of view (FOV) of the device under test (DUT), in front of a static, uniform neutral background, which can be reflective (front-illuminated) or transmissive (back-illuminated). During image capture, the DUT should be mounted on a tripod or other fixed fixture.
The LED light source should have adjustable frequency, duty cycle, and intensity. The “ON” and “OFF” reference light sources must remain within the DUT's FOV throughout the test. The “ON” reference light source must be identical in design and physical characteristics to the test light source, but driven by a constant current source with matched luminance. The “OFF” reference light source must be identical in design and physical characteristics to the test light source, but remain turned off in all test scenarios. The luminance of the background should be configurable.

“Reflectance Flicker”:
The test target should be a uniform neutral background illuminated by a modulated light source. It can be a transmissive target with an integrated modulated light source, or a reflective target uniformly illuminated by a modulated light source. The deviation in illumination uniformity should not exceed ±5% to ensure the accuracy of tone measurements. For cameras with fisheye lenses, curved or bowl-shaped targets may be required.
Since banding artifacts are also closely related to the camera's exposure time, for cameras without manual exposure control (black-box cameras), it is also recommended to use an unmodulated light source to vary its exposure time.

4. Test Methods and Requirements
Test Procedure:

• The video sequence must be long enough to ensure the capture of a complete phase cycle;
• Throughout the test, the maximum dynamic range setting of the camera must not be changed, and overexposure in the captured area must be avoided;
• The frequency of the pulse width modulation must not be an integer multiple of the camera's frame rate, otherwise phase coverage may not be uniform enough;
• In actual testing, the selection of the LED frequency range should consider specific application scenarios. For example, the flicker frequency of some light sources is intentionally set within the human visible range, such as turn signals and emergency lighting. The ability of the DUT to correctly identify the occurrence of such flicker should also be included in the test scope.

Test Requirements:
“Illuminant Flicker”:
Test Target Requirements: The PWM light source should be located near the center of the camera's field of view (FOV) to minimize lens shading effects. A diffuser must be installed in front of the light source. The diffuser must have a uniform surface (peak variation ≤10%) and exhibit matte Lambertian characteristics to minimize specular reflection, and holographic diffusers are prohibited. All target areas should be made of the same material, and the image area of all regions should be equal.
Region of Interest (ROI) Selection: Minimize the number of rows: To accurately capture the flicker phase, it is recommended to minimize the number of measured rows, ideally measuring only a single row of pixels (especially for rolling shutter sensors).
Maximize the width: To reduce the impact of image noise, it is recommended to set the ROI width to the maximum allowed by the target area to increase the number of effective pixels.
Reference Light Source Setup (to ensure measurements are within the camera's dynamic range): Two reference areas must be set up next to each PWM flickering area:
Reference “OFF”: Same material as the PWM area, but unpowered (as shown by 'c' in the figure).
Reference “ON”: Same material as the PWM area, but driven by a constant current (always on, no flicker). Its luminance has two modes: 1. Matched to the equivalent luminance of the PWM (as shown by 'b' in the figure); 2. Maximum luminance of the PWM (100% duty cycle) (as shown by 'a' in the figure).
Test Strategies for Different Types of Cameras:
Configurable Cameras (fixed exposure/ISP): A sequential test method can be used, i.e., capturing the “ON”, “OFF”, and “flickering” states sequentially.
Black-box Cameras (auto exposure/ISP): The sequential test method may yield unreliable results due to the camera's automatic adjustments. Therefore, a synchronous test method is recommended, where the “ON”, “OFF”, and “flickering” targets are placed in the scene simultaneously and captured in a single shot.
Additionally, note that: the size of the target light source should be limited to within 5% of the image height to minimize the impact on the camera's automatic algorithms; the scene should include a constant background light source with a luminance no less than the reference “ON” light source to stabilize the scene's dynamic range; after switching between different test modes, a stabilization time must be allowed for the camera's automatic adjustment functions to stabilize.

“Reflectance Flicker”:
The test equipment should capture a series of videos of the test scene at a fixed frame rate, with the light source modulated at different frequencies in each video. The duration of the captured videos should be at least a few seconds, and long enough to observe the changes in banding artifacts. When the modulation frequency of the light source approaches an integer multiple of the test equipment's frame rate, the temporal modulation will slow down. When the modulation frequency is exactly an integer multiple of the frame rate, no temporal flicker effect will occur. With the use of a rolling shutter, stationary bands will appear in the image. The most difficult case to test is when the light modulation frequency differs from these points by only a fraction of a Hertz, as verifying that it is indeed changing may require a considerably long video.

5. Key Performance Indicators (KPIs)
“Illuminant Flicker” KPIs:
Flicker Modulation Index (FMI)
$$ FMI = 100 \times \frac{x_{\max} - x_{\min}}{x_{\max} + x_{\min}} $$ For a given time series s(n), where: FMI is the Flicker Modulation Index; $x_{\max}$ is the maximum measured signal; $x_{\min}$ is the minimum measured signal. s(n) is a temporal signal generated by averaging the ROI values in each frame of the test video sequence.

Flicker Detection Indicator (FDI)
The FDI (Flicker Detection Indicator) is a quantitative metric used to evaluate the camera system's ability to suppress or capture the flicker of pulsed light sources (such as PWM-driven LEDs). It is a probability value (up to 1.0) that reflects the proportion of frames in the captured video where the flickering light source can be clearly distinguished from the background reference luminance. A higher FDI indicates that the camera is more likely to “see” the flickering light source (good flicker detection capability); a lower FDI indicates that the camera has difficulty capturing the flicker (poor flicker detection capability, which may cause the light source to appear constantly on).

Measuring FDI requires following these standard steps:
1. Define the minimum contrast threshold (τ): This is the threshold value to determine whether the light source in a frame has been successfully “detected”. The standard recommends using Weber contrast, with the formula: (measured signal - reference background signal) / reference background signal. The specific threshold is determined by the application scenario. If no specific requirements are defined for a particular camera application, the IEEE 2020-2024 standard recommends a minimum Weber contrast of 20%.
2. Set up the test environment: Determine the LED flicker frequency and duty cycle to be tested. Define and report the dynamic range of the test setup (i.e., the luminance difference between the flickering light source and the reference “OFF” state).
3. Collect data: a. Capture images or video of the reference “background/OFF” state to determine the reference signal level (x_refoff). b. Capture a video sequence of the pulsed light (flickering LED).
4. Calculation and Analysis:
For each frame in the video, calculate the pixel average of the flickering light source area (x_meas), and calculate the Weber contrast with the reference signal level (x_refoff). Count the number of frames in the entire video sequence where the Weber contrast exceeds the preset threshold (τ).
FDI calculation formula:
FDI = (Number of frames with contrast exceeding the threshold) / (Total number of frames in the video)
Expressed mathematically:
$$ FDI = Prob\left(\frac{x_{\mathrm{meas}} - x_{\mathrm{ref\,off}}}{x_{\mathrm{ref\,off}}} \geq \tau \right)$$ Where $FDI$ is the Flicker Detection Indicator; $Prob$ is the probability; $x_{\mathrm{meas}}$ is the average luminance of the measured flicker light source at different frequencies; $x_{\mathrm{ref\,off}}$ is the average luminance of the measured OFF light source at different frequencies; $\tau$ is the flicker contrast threshold.

Modulation Mitigation Probability (MMP)
The Modulation Mitigation Probability (MMP) is a metric that, for a given camera, LED frequency, duty cycle, and phase angle, measures the probability that the camera reproduces the light source signal within a defined signal range relative to the reference signal level in the captured images. The reference signal level is the average of the measured light source signal levels. The threshold δ is a key parameter that defines the acceptable boundaries of the image modulation level. The selection of this value depends on the specific application requirements of the camera. If there are no special requirements for the application, this standard recommends a default value of δ = 0.1. Its calculation formula is as follows: $$\text{MMP} = \text{Prob}\left[ \overline{x_{\text{ref}}}(1-\delta) < x_{\text{meas}} < \overline{x_{\text{ref}}}(1+\delta) \right]$$ Where $\text{MMP}$ is the Modulation Mitigation Probability; $\textbf{Prob}$ is the probability; $\textbf{$x_{\mathrm{meas}}$}$ is the measured flicker signal level; $\textbf{$\overline{x_{\mathrm{ref}}}$}$ is the average signal level of the imaged light source; $\textbf{$\delta$}$ is the acceptable threshold level.

Flicker Beat Frequency (FBF)
FBF refers to the frequency of the modulated light source captured by the camera system. It is the modulation frequency within the video sequence, not the frequency of the physical light source itself. If the camera frame rate and light source frequency are known, FBF can be calculated using the following formula: $$FBF = \min\left[ \text{mod}(f_{\text{scene}}, f_{\text{camera}}), \text{mod}(-f_{\text{scene}}, f_{\text{camera}}) \right]$$ Where \( FBF \) is the Flicker Beat Frequency; \( f_{\text{scene}} \) is the light source frequency; \( f_{\text{camera}} \) is the camera frequency.

“Reflectance Flicker” KPIs:
Flicker Modulation Index (FMI)
$$FMI = 100 \frac{x_{\text{max}} - x_{\text{min}}}{x_{\text{max}} + x_{\text{min}}}$$ Where: FMI is the Flicker Modulation Index; $x_{\max}$ is the maximum measured signal; $x_{\min}$ is the minimum measured signal. The calculation principle of this metric is essentially the same as that of the Illuminant Flicker FMI.

Dark Band Height
The height of the “dark band” (i.e., the number of image rows of the dark band) applies only to rolling shutter sensors. It is measured from the midpoint of the contrast transition.
The dark band height can be reported in the following units: a. Number of image rows; b. Percentage of image height; c. Dimensions on the output display.

Light Band Height
The height of the “light band” (i.e., the number of image rows of the light band) applies only to rolling shutter sensors. It is measured from the midpoint of the contrast transition.
The light band height can be reported in the following units: a. Number of image rows; b. Percentage of image height; c. Dimensions on the output display.

Flicker Beat Frequency (FBF)
The temporal frequency of light modulation, expressed in Hertz (Hz). For rolling shutter sensors, this is the rate at which the bands move across the image. For global shutter sensors, this is the frequency at which the image brightness changes over time. By convention, a positive number indicates that the bands move upwards or to the right in the image from the observer's perspective.

Edge Rise Distance
The 10% to 90% edge rise distance between the dark band and the light band. It is recommended to measure the edge rise time in the Y, R, G, and B color channels, or the camera's linearized raw output.

6. Yanding Equipment Support
MLB-HMC ADAS Camera Comprehensive Tester
This equipment is specifically designed for ADAS camera image quality testing, capable of evaluating image quality in complex lighting environments and under various motion conditions. It can be used for CPI, Flicker, and Motion testing of ADAS cameras. Its dedicated software control system ensures a high degree of consistency in test conditions, allowing for efficient measurements by importing scripts after a single setup, making it suitable for horizontal performance comparisons of cameras.

RIQA-ADAS Image Quality Analysis - ADAS Module
RIQA-ADAS is a professional image quality analysis software developed specifically for autonomous driving and in-vehicle cameras. The algorithms of its Flicker in/Flicker out modules are based on the IEEE 2020-2024 standard, capable of quantifying the flicker mitigation capability of camera modules.
Flicker in

Flicker out