When HDR looks washed out, it’s often because the HDR display does not appropriately limit and compensate for light that is added to the image from the display’s backlight or glare. This has nothing to do with Windows or macOS behaviour and is not because of how sRGB content is incorporated into an HDR image.
Introduction
If you’ve researched or used external HDR displays, especially when connecting one to a computer, you probably know the problem of HDR mode appearing washed out and lacking saturation. This problem is typically caused by the display not limiting or appropriately compensating for light that is inadvertently added to the image. This light typically comes from the display’s backlight or glare. Interestingly, this problem is not common with sRGB SDR displays, so we’ll take an in-depth look into why that is in the latter part of this post.
An HDR display’s image may also appear desaturated or low contrast because of the display’s highlight compression or gamut mapping, for example, but these issues vary between displays and will not be explored in this post.
Backlight Bleed-Through
Let’s start by visually comparing the difference between SDR and HDR mode on the Asus PA279CV. This is an LCD monitor that looks great in SDR mode and looks very washed out in HDR mode. Slide the handle from left to right to compare the two photos:
SDR mode photo | HDR mode photo
In the above photos, the display is receiving a video signal that includes pure black in the top left corner of the image… But it’s clear that this “pure black” part of image appears quite a bit brighter when in HDR mode. The reason for this difference is because of the brightness of the LCD’s backlight: the backlight’s brightness is increased when in HDR mode, which means that more white light bleeds through the LCD and raises the black level of the image in HDR mode. Because light is perceptually non-uniform to humans, the colour rendering of darker or saturated colours is affected more than bright and unsaturated colours when light is added:
I measured the luminance on this display to find that a pure black image in an extremely dark room produces 0.32 nits in HDR mode and 0.07 nits in SDR sRGB mode. The following charts show the luminance in HDR mode that I measured using a Konica Minolta LS-100 luminance metre. The left chart uses linear scale and the right chart uses a base 2 logarithmic scale, also known as exposure stops. When looking at the linear scale, it may, at first, appear that the error decreases slightly with lower luminance input signals, but as demonstrated above, the human eye is more sensitive to differences in low relative luminance light than high relative luminance light. Because of this, the nonlinear logarithmic scale on the right side presents a rough approximation of the error perceived by a human: dark values are impacted most from light that is bleeding through from the backlight.
To address this issue, newer HDR displays will limit this added light by dimming the backlight on different parts of the screen or using technologies that have no backlight at all. The problem is more nuanced with displays that implement “local dimming” of the backlight because bright and saturated colours should have high luminance in one colour channel, but very low or zero luminance in other colour channels; the backlight can often not be dimmed for only one or two of the three colour channels.
The following photo demonstrates the local backlight dimming of a 14-inch Liquid Retina XDR Nano-texture display of a 2025 MacBook Pro (Apple M4 Pro chip). I have physically blocked out the middle part of the image to reduce flare and allow us to see the white backlight bleeding through what should be perfectly black pixels on the left side of the photo:
Notice that white light that bleeds through the LCD for areas that have bright colours, regardless of whether it is only one colour channel that should be very bright. This means that, although the backlight dimming helps with dark values, the same problem of bright and saturated colours becoming desaturated may still happen with this type of locally dimmed display. Later in the post, we’ll look at how this specific MacBook Pro display addresses this issue.
Glare
Another common source of problematic light is glare. In the context of display behaviour, glare is light that been added to the presented image, but has a source that is external and unrelated to the display technology. This light may have either reflected off the surface of the screen (“specular” light) or refracted into the subsurfaces of the screen, bounced around, and finally refracted outward to the observer (“diffuse” light). This “diffuse” light is especially important for colorimetry because it strongly influences what a person describes as the colour of a surface and is not easily understood by an observer’s visual system as reflections on the screen that are unrelated to the intended image.
We sure have come a long way since the days of CRT monitors—it’s easy to see the impact of anti-glare technologies on many newer screens, especially LCD displays.
The Asus PA32UCDM QD-OLED display has a whopping 0.541 nits of glare with the lights turned on in my office, mostly due to not having a polarizing filter. This amount of light that is added to the image is more significant than the backlight bleed-through of the LCD from the beginning of this post! Even though this QD-OLED display is not impacted by light bleeding through from a backlight, one would expect the HDR mode of this display to appear very washed out and lacking in saturation due to this glare. But surprisingly, this display produces a rich and saturated image with a similar appearance to the other LCD display in SDR sRGB mode. Let’s look into why this is…
Choosing a Compromise
Computer displays, such as those found in smartphones, tablets, laptops, and used with many desktop computers, use a common philosophy: When a colour at a relative luminance is requested of the display, it is the display’s responsibility to present it as accurately as possible according to CIE colorimetry and how the expected viewing environment will interact with the display’s characteristics. This is the same for the HDR10 standard, which describes a video signal that requests colours of an absolute luminance be presented by the display.
Ideally, display hardware should present colours accurately, no matter how dark or saturated they are. But, as demonstrated above, this is often impossible with the limitations in consumer hardware. When it is not possible to present that colour accurately, due to light that has been added to the display from backlight bleed-through, glare, or other sources, a compromise must be made. Here are the two most basic and common compromises:
- Ignore light that has been added to the display
- Benefit: all image detail is retained and visible
- Problem: image appears washed out and desaturated; colour accuracy is decreased for dark or saturated colours
- Compensate for light that has been added to the display
- Benefit: colour accuracy is the best possible according to the CIE standard observer, given the constraints
- Problem: colour information and detail in the darkest and most saturated parts of the image is entirely lost or becomes difficult to see
There may exist a third approach that achieves the benefits of approach 1 without the problem of a washed out appearance, but the problem of decreased colour accuracy would remain because it would be necessary to deviate from a stimulus that is mathematically equivalent to the input signal’s tristimulus according to the CIE standard observer.
The first approach of “ignoring added light” is used by the Asus PA279CV LCD that was discussed in the Backlight Bleed-Through section of this post. The remainder of this post explores the way that typical computer and integrated displays implement the second approach, both with HDR and sRGB SDR.
Compensating for Added Light
The simplest way to compensate for light that has been inadvertently added to an image is to subtract the same amount of light from the image before displaying it.
This technique results in a loss of detail in dark or saturated colours, but does produce accurate rendering of colours that are possible to present when this exact amount of white light has been added to the image from sources like backlight bleed-through and glare. This is inherently a bit of guesswork because the display manufacturer cannot be sure of exactly how much glare may be added to the image or what the colour of this light may be. But the manufacturer does know the exact glare characteristics of their display, so they can take a reasonable guess based on the environment that they expect their users to view the display. Further, the manufacturer may expose this compensation amount to the user as a setting on the display or integrate a light sensor that allows this adjustment to be configured automatically.
The simple clipped subtraction approach shown above must be avoided in practice because this will produce visible bands in gradients that reach lower luminance values in one or more colour channels. Since I’m not actively writing display firmware and hardware calibration, I will not be doing the math to figure out the exact piecewise function that has a suitable smoothness and a slope of 1.0 at the crossover point, but don’t hesitate to reach out if that’s something you have a need for and I may be able to provide this. In practice, this would likely be implemented with a power function for the nonlinear segment because this approximately matches the human eye’s ability to differentiate shades at lower relative luminance.
Regardless, here’s a rough sketch of what I expect this sort of smoothed compensation technique would look like when measured:
It’s possible that this smooth approach allows more detail to be preserved in dark colours of an image than simply clipping, but due to the raised black level and compression of the darkest values, details cannot be preserved at the same fidelity as an ideal display that does not have any unwanted light added to it. It is, as always, a compromise that must be made for imperfect hardware and viewing environments.
One final note: this simple compensation technique assumes that the colour of the undesirable white light that has been added to the image is the same colour as the display’s white point. If it isn’t, then a more accurate adjustment would be to compensate by different amounts for each of the colour primaries. This could be implemented with a sensor that measures the colour of external light that is interacting with the display.
Compensation Techniques in Practice
Enough theory, let’s take a look at some measurements of HDR displays that do not appear washed out to see how they have solved this problem.
Asus PA32UCDM QD-OLED
As I mentioned in the Glare section, the Asus PA32UCDM QD-OLED display has a very high glare characteristic, but doesn’t appear washed out. The following are measurements taken without glare (in an extremely dark room) and with glare (in a typical office room with the overhead lights turn on and no natural light).
This display compensates for added light and the result is an appearance that does not look washed out, even when there is substantial ambient light that results in around 0.5 nits of glare. It’s easy to tell from the chart that there is too much compensation applied and the display will look too dark, even with 0.5 nits of light added from glare. And I can confirm this with my own eyes.
This display also provides a “Black Level” setting, which defaults to a value of 50. The previous chart shows this factory default behaviour, but let’s take a look at how the monitor behaves with different Black Level settings:
This display’s Black Level setting controls how much compensation for added light is applied to the image before it is presented. If you are in a pitch-black room, a Black Level of around 70 is likely appropriate. In my office with the overhead lights turned on, a Black Level of close to 60 is likely reasonable. I am uncertain of why the factory default setting of 50 was chosen, but maybe this would produce the best image in a bright daylight environment, at the cost of not being able to see detail in dark or saturated colours.
2025 MacBook Pro Liquid Retina XDR
Next, let’s take a look at the 14-inch Liquid Retina XDR Nano-Texture Display built into the 2025 MacBook Pro. As discussed in Backlight Bleed-Through, this display has a local backlight dimming feature that limits the amount of light that bleeds through the LCD when dark colours are presented, but still needs to compensate for a very bright backlight when bright and saturated colours are presented. Here are the measurements, both without and with glare from my overhead office lighting, when using a device brightness setting that maps reference white to 140 nits:
I expect that this amount of compensation for added light helps ensure colour accuracy in very bright and saturated colours, at the cost of darker colours appearing too dark and saturated on this display.
Next, here is the same display with the device brightness set to a reference white luminance of 600 nits:
We can see the amount of compensation scales with the device brightness setting. Glare is extremely limited, even in brighter environments, because of the nano-texture treatment on this display. So I am unsure of whether this extreme darkening at higher reference white luminance is due to a limitation in technology and/or attempting to maintain similar relative behaviour in dark or saturated values.
It’s hard to say if this MacBook display is applying too much compensation for added light at lower reference white luminance due to the complexities of a variable brightness backlight. But either way, its compensation is nowhere near as much as the Asus PA32UCDM unless the device brightness is set very high. I have also measured the behaviour with True Tone enabled to find the curve is similar, but scaled to a different reference white and white point luminance.
SDR and sRGB
Displays in SDR mode also have some light added to their image from sources such as backlight bleed-through and glare, even if it might be less than certain displays in HDR mode. For example, when I have the overhead lights turned on in my office, a pure black colour on the Asus PA279CV LCD display in SDR sRGB mode measures as 0.217 nits. So why doesn’t this image appear somewhat washed out?
In my previous post, I described how there appears to be no official documentation that describes a specific benefit or functional purpose of the mismatch between the 2.2 gamma reference display and the (piecewise) encoding implementation. But even an untrained eye can see the difference between an image that has this mismatch and one that doesn’t: the mismatch causes dark colours to appear darker and saturated colours to appear more saturated. So even if the standard does not explicitly state it, is there a practical and functional purpose to this mismatch?
The sRGB standard provides reference conditions, both for the display and the viewing environment:
- Reference white luminance: 80 nits
- Display offset (bias/lift): 0.0
- Power function exponent (gamma): 2.2
- Veiling glare: 0.2 nits
The display offset and veiling glare are important in defining the luminance that should be presented by an sRGB display. A 0.0 offset with a 2.2 gamma means that the display should emit light of 0.0 nits when given a pure black signal and that the electrical behaviour of the display should match a simple 2.2 power function. A veiling glare of 0.2 nits means that 0.2 nits of white light must be added to the presented image to produce a correct stimulus, according to these reference conditions. (The IEC definition for “veiling glare” is the same as I have used for “glare” in this post.)
Let’s take a look at the expected luminance of an sRGB reference display with and without this 0.2 nits of glare added:
The mismatch between the piecewise sRGB encoding and the 2.2 gamma used by the reference display provides compensation for the 0.2 nits of added light from glare, that is defined by the reference conditions, by taking advantage of the common 2.2 gamma CRT monitors that were widely available at the time the standard was introduced.
Although intent was omitted in the sRGB standard, the mismatching 2.2 gamma acts as a compensation for added light in practice. In section 5.2, the sRGB standard states that transformations from nonlinear piecewise-encoded sRGB code values to linear CIE 1931 XYZ values should use the inverse piecewise functions rather than a 2.2 power function. Now these instructions make sense:
These CIE 1931 XYZ values represent optimum image colorimetry when viewed on the reference display, in the reference viewing conditions [that describe 0.2 nits of glare], by the reference observer, and as measured on the faceplate of the display, which assumes the absence of any significant veiling glare [that is greater than the 0.2 nits reference amount].
(Emphasis and [parenthesis] mine)
By using the inverse piecewise sRGB function to decode sRGB code values to a different colour encoding, you will get an image that is similar to the one presented on an sRGB reference display with 0.2 nits of glare because this glare and the 2.2 gamma mismatch will effectively cancel each other out.
This compensation works well for added light that is about 0.25% (0.2 nits / 80 nits) of the reference white luminance. LCD displays, which have lower glare characteristics than a CRT, exhibit bleed-through of a backlight that is functionally similar to the glare discussed here; the mismatch can similarly act as a way to compensate for the bleed-through of an LCD display’s backlight or light from other sources. Finally, because this compensation for added light is fixed at 0.25%, it is expected that added light will increase proportionally as reference white luminance increases. This is often the case for many types of displays as response to a brighter viewing environment: the amount added light increases either due to increased glare from a brighter viewing environment or from an increased backlight brightness.
sRGB in Practice
Again, enough theory. Let’s take a look at the measured luminance of the Asus PA279CV LCD from the beginning of this post in SDR sRGB mode with the overhead lights turned on in my office:
It just so happens that the backlight bleed-through plus glare from the overhead lights adds up to 0.217 nits—perfect for the added light compensation of the sRGB standard and this 2.2 gamma SDR display.
sRGB in HDR
Since we now know that the practical function of the sRGB 2.2 gamma mismatch is to provide compensation for light added by sources such as glare and backlight bleed-through, we can better reason about how sRGB content should be encoded into an HDR signal.
Unfortunately, there exist HDR displays that limit and compensate for light that has been added to their image and displays that ignore and do not limit this added light. All content, including HDR content, will appear washed out on HDR displays that ignore and do not limit this added light. To address the washed out appearance on these types of displays, some sort of added light compensation must be applied to the entire HDR signal sent to the display. Using a 2.2 power function to “decode” sRGB content for integration in an HDR signal will, at best, only correct the appearance of sRGB content, leaving HDR content appearing washed out.
For displays that implement their own ways of limiting and compensating for added light, we must use the inverse piecewise sRGB encoding function to attain linear RGB values for use in the HDR signal. Using a 2.2 power function to “decode” sRGB content for this type of HDR display will double-up compensation for added light, resulting in sRGB content that is too dark and over-saturated.
HDR Behaviour in Windows and macOS
There is a common misconception that HDR on Windows is “bad” and sometimes these opinions will go even further to say that this problem is specific to Windows and does not happen with macOS. Let’s compare a display in HDR mode using Windows and macOS with an sRGB SDR app and see if there are any differences:
Windows and macOS produce an identical image in HDR mode: sRGB apps are decoded using the inverse piecewise sRGB function on both Windows and macOS, which produces correct behaviour on HDR displays that implement their own methods for limiting and compensating for light that is added to the image from sources such as glare or backlight bleed-through.
Appendix: Measurements
Here’s the full set of data and measurements from my LS-100 luminance metre. This spreadsheet contains some additional data beyond what was presented in this post, such as luminance measurements of P3 primaries on the MacBook display and a comparison of luminance behaviour between 60 Hz and 240 Hz mode on the Asus QD-OLED display.