Reproducible quantification of display sparkle

Display sparkle is a high-frequency, random luminance and chromaticity variation that appears on top of the regular pixel pattern of a display. It is associated with displays that carry an anti-glare layer (AGL) — a common cover-material option for displays used in bright environments. AGLs are widely used in the automotive sector and in other applications where ambient-light readability matters, but not every automotive or outdoor display uses one.

Where an AGL is present, it is the physical cause of the effect. The metric used to quantify it is the sparkle contrast defined in IEC 62977-3-9.

The anti-glare layer scatters light. That is its purpose in reflection: incoming ambient light is scattered away from the viewer so that mirror-like reflections on the display surface are suppressed and contrast in bright environments improves.

The same scattering also affects the light that the display emits through the AGL. The microstructure of the layer refracts and scatters the light from individual subpixels in slightly different directions. What arrives at the viewer is a random, high-frequency luminance variation on top of the regular pixel grid. That variation is what we measure as sparkle. Sparkle strength depends on the AGL design (foil vs. glass, roughness, layer thickness, distance to pixel layer) and on the display behind it (subpixel layout, PPI).

Sparkle measurements are sensitive to several setup conditions: where the focus is set, how finely the camera samples the display, what aperture the lens uses, and which algorithm is used to separate the sparkle signal from the regular pixel pattern. If these conditions are not controlled, two labs measuring the same panel can read very different values.

IEC 62977-3-9 acknowledges this in two ways. It requires every measurement to document the relevant setup conditions, and it permits several different image-processing and capturing methods to separate the sparkle signal from the pixel matrix. The TechnoTeam procedure addresses each of these conditions in turn: a distance focus scan handles the focus position, Fourier downsampling handles the sampling resolution, an in-house aperture window keeps the angular aperture in a reproducible range, and the chosen separation algorithm is a frequency filter that stays compatible with BlackMURA setups.

Sparkle contrast is the metric used to quantify display sparkle. It is expressed in percent and defined as the standard deviation of the luminance divided by the mean of the luminance. For the frequency-filter method, it is calculated on the frequency-filtered image rather than on the raw capture and within a defined region of interest.

A captured luminance image of the display is transformed into the frequency domain. There, two contributions are filtered out: the periodic component from the pixel grid and the low-frequency component from large-scale luminance variations such as backlight gradients. The image is then back-transformed to the spatial domain, and the sparkle contrast is evaluated on the filtered image. What remains is the high-frequency signal that corresponds to the perceived sparkle.

The frequency-filter approach is one of several separation methods that IEC 62977-3-9 permits. TechnoTeam selected it because it works under similar sampling and alignment conditions as BlackMURA, so a single measurement cell can run both measurements. To make the result comparable across different cameras and lenses, the inverse Fourier transform additionally suppresses frequencies above a common reference — a step that removes the dependence on the actual sampling resolution within the validated measurement window.

The random sensor noise of an imaging luminance measurement device (ILMD) looks similar to sparkle in the captured image: both are high-frequency and random. Once the periodic pixel grid has been filtered out, sensor noise can therefore contribute to the measured sparkle contrast and would falsely increase the reported value. The effect is small for displays with strong sparkle, where the sparkle signal dominates, but it can become significant at low sparkle levels — close to the level of a panel without any AGL.

The countermeasure is averaging. Several ILMD captures of the same image are taken and combined before the sparkle contrast is calculated. The sparkle signal stays constant under this averaging, while the random noise component drops by a factor of √N for N captures, so a small number of captures already reduces the noise contribution to a level well below the sparkle signal.

The method was validated against subjective expert ratings during its development. Automotive-display quality engineers rated a set of AGL samples on a low, medium and high scale without knowing the measured values, and the sparkle contrast values then followed the visual ranking with a clear monotonic relationship across the three categories. The procedure has since been verified in a round-robin experiment across three independent laboratories.

The acceptance threshold for a "good" sparkle value is set by the OEM or by the application. Also IEC 62977-3-9 does not define one.

No. Our procedure was designed so that the AGL stays on the display, which is what makes the measurement usable across the full lifecycle: AGL development, supplier qualification and incoming inspection on the assembled module. The same procedure applies to coated cover glass on a reference display and to a finished module, and both glass-based and foil-based AGLs are covered.