LED Displays 102: Image Pipeline

LED Displays: Image Pipeline

An important aspect of all display technologies is how they reproduce image content, that is, the electronic signal that represents the image to be viewed on the display surface. We’ll start by looking at how a grayscale image is made, and then add color to that.


At their most basic, emissive raster images are made up of pixels in a regular grid each having varying intensities. When viewed in aggregate, the eye fuses all of these individual pixels and their different intensities into what appears to be a continuous image.

Display Brightness

One of the fundamental properties of LEDs is that they can either be on or off. Unlike an incandescent light source, they don’t have an analog ‘brightness’ that can be adjusted. To achieve what appears as differing intensities, LEDs are flashed on and off at an extremely high refresh rate which the human vision system temporally integrates into a an apparent continuous brightness. By varying the duty cycle of the flashing pattern, different luminance levels can be created.

Bit depth decimation

Since an LED can only be on or off, it follows that its maximum intensity is achieved by leaving it turned on as long as possible per frame time. Conversely, what happens if we want to lower the maximum intensity of the display after the display's electronic properties have been defined, and the display has been produced? To do this, we must adjust the intensity digitally, by reducing the duty cycle.

$$ \begin{equation} a^2+b^2=c^2 \label{eq:pythagoras} \tag{1} \end{equation} $$

Referring to Eq. $\eqref{eq:pythagoras}$.

100% Maximum => 12.5% maximum $100/2^3$ = 3 bits lost. 16 bit maximum is now 13 bit maximum. Related to this, the smallest difference in luminance is always the same, regardless of what the maximum luminance of a display is set to. $$1 bit => 1500cd/m^2 * \frac{100\%}{2^{16}} => 0.02 cd/m^2$$ From this, we can see that applications that require lower maximum intensities are harmed by the numbers race of LED manufacturers in their quest for ever larger luminance ratings. If you know in advance the maximum luminance needed for your application, you’re much better off buying a display with 1x headroom than 10x headroom.

Full Range vs. Limited Range

The bit depth available to a digital image is made even more complex by full range vs. limited range codings. In order to simulate the way analog signals were handled, a limited range encoding was devised during the transition from analog to digital signalling. This coding uses a subset, or limited range, of the full bit depth available to reproduce images to better map to the analog signals that used a subset of the analog range to account for analog signal behaviors such as overshoot and ripple. Data outside of this range is used to maintain compatibility with analog equipment, but is expanded and thrown away when reproduced on a fully digital display technology.

HDR and Bit Depth


White Point

Color Spaces

Accuracy vs. Precision

Uniformity compensation vs. Calibration

Standardized Color Spaces and Content Management

Color space standardization allows multiple devices to display images with the same perceived color. This allows an image that is prepared or manipulated on one device to be viewed on another, and have it appear as expected by the viewer.

Gamma, Transfer Functions, and Color Grading

OETF/EOTF, OOTF Rendering Intent

Thermal Color Shift

Synthetic vs. Sensor-based Image Content

RGB vs. YCbCr

Signal Compression

Chroma Subsampling

Progressive and Interlaced Formats