Energy Efficient Control of LED Backlight TV Displays and Ambient Light

local backlight TV

Based on image data the LED backlight values are dimmed and thereafter the LC pixel values are adjusted


DTU Fotonik and Bang & Olufsen have in a research collaborative effort developed and demonstrated energy efficient advanced control of TV displays with LED backlights. This includes the utilization of measurements of the ambient light at the time of viewing the TV.

The goal has been a simultaneous reduction of energy consumption and image quality improvement in typical TV viewing in a living room.

The collaborative research has developed and demonstrated energy efficient TV platforms with local dimming of LED backlight, including ambient light sensors and measurements. Demonstration and testing has been conducted in an environment with controlled ambient light to demonstrate the light aware control and energy efficient solution. The research is based on a novel frame work for energy optimal, high quality LED backlight display control

The research has included demonstration of energy efficient control of LED backlight LCD displays for two different architectures including measurements of ambient light, namely edge-lit and direct-lit backlight. The research has also included HDR (High Dynamic Range) aspects and display. For high quality TV material, 30% relative image dependent energy savings by local dimming of the display. Taking measurements of the ambient light into account savings up to a factor of five can factor in on top for up to a combined factor of seven savings.

The LED backlight TV is dominating the market and the backlight accounts for most of the energy consumption. Thus efficient solutions to control the LEDs provide significant potentials for saving energy in households worldwide.

The research has been supported by national funding under the EUDP and DSF programs, with energy efficiency and Green ICT as topics.


The goal of the research is simultaneous reduction of energy consumption and truly im-proved experienced image quality for typical TV viewing. The approach in the EUDP project, to achieve the goal, was adaptive control of LED backlight including the use of ambient light measurements. The adaptive algorithms for backlight control are based on analysis of images and use of information of the ambient light measurement.


The novel approach is based on energy aware optimization. The project worked with and demonstrated LED backlight control for two different backlight architectures and for HDTV and UHD. The technique also included as an important input measurement of ambient light for optimal energy efficiency. For high quality TV material relative image dependent energy saving of 30% for the display was achieved. Taking the ambient light into account energy saving of up to seven times was demonstrated. It was shown that including the ambient light plays a crucially important role in energy efficient backlight control with high visual quality.


1. Green TV technology gives a good black.


Bang & Olufsen and DTU Fotonik developed and demonstrated energy efficient advanced control of TV displays with LED backlights. The project developed and demonstrated energy efficient TV platforms with local dimming of LED backlight. The main platform for the project was an experimental programmable platform established jointly at both DTU Fotonik and Bang & Olufsen.

The results of the project included demonstration of energy efficient control of LED back-light LCD displays for two different architectures including measurements of ambient light.

The LED backlight TV is dominating the market and the backlight accounts for most of the energy consumption. Thus efficient solutions to control the LEDs provide significant potentials for saving energy in households worldwide. Bang & Olufsen has used ideas on the backlight dimming from the collaboration in their newest TV set, BeoVision Avant.

The energy goals of the project were achieved including providing energy optimal, high quality LED backlight display control.


Flat panel displays have within a few years taken over to dominate the TV market. The LCD panels today constitute the bulk of the market. Despite their prevalent success, the LCD panels can still be improved in terms of energy consumption and video image quality. The recent generation LCD panels the so-called LED TV with LED backlight and local backlight dimming, offer the potential for significant energy savings and at the same time improved image quality. The local backlight dimming displays imply a redefinition of the video signal at time of display; instead of having one backlight, it is possible image dependently to locally reduce the backlight and thereby save energy. The image is reproduced as a combination of the LCD element transmission and the local level of the backlight. The video signal processing including the control of the local backlight with the goal of simultaneous reduction of energy consumption and truly improved image quality for typical TV viewing has been a subject of our studies. The LED backlight is illustrated on the figure starting from the LED backlight, which is diffused and attenuated by the LC elements for form the resulting image.



Displaying the image on the BeoVision Avant.using local dimming with direct backlight (85”) and edge-lit backlight (55”). The backlight emitted by the LEDs (a and b) is diffused (c and d), and the LC elements adjusted to form the resulting image (e and f).


A novel quality-energy saving frame work for backlight dimming has been developed. This was further developed in form of perceptual domain quality and fast versions of the algorithm. Also control of power and quality over time in video sequences was developed. The algorithms were tested on an experimental platform developed in the project and visual quality testing was performed. The basic problem is that LC elements in LCD panels can not block the light complete, thus light leaks out in dark parts of an image thus reducing the image contrast. The backlight dimming attenuates this but at the risk of not providing enough light for the bright areas. The challenge is to strike the best balance. This has been formalized and solved mathematically also including the energy.  


To illustrate the effects, the figure depicts the effects of dimming vs. no dimming. (The image is composed of two photos of a Bang & Olufsen screen displaying the same image but with and without dimming. The left part is from the set-up with no dimming and the right part from the set-up with dimming.) The dimming directly gives energy savings.


Illustration of dimming vs. no dimming. In the right part local dimming (direct dimming) is applied, whereas in the left part full backlight is used and the leakages of light through the LC pixels are clearly visible.


The work has been carried out in three research projects as a collaborative effort between DTU Fotonik, and Bang & Olufsen. In parts of the research, also with participation of TekPartner and McMaster University, Canada. Experimentally we have been working both with an HDR (High Dynamic Range) monitor from SIM2 and an experimental platform based on a TV-set, established in collaboration with Bang & Olufsen and TekPartner.


2. Local backlight dimming research


A major activity has been to establish a broad understanding of the problems from material science over video and real-time play back with control of the screen to perception aspects.


A general model of backlight dimming, display and perception has been established and further developed throughout the project. Optimal backlight has been derived in a general framework allowing for further optimization including feed-back and in-sights to be gained in perception based testing. Also a fast version of the backlight control has been developed towards real-time implementation.


To display and for subjective visual quality assessment, controllable backlight platforms with real-time play-back have been established. A direct-backlit HDR display has been studied and optimal algorithms tested. An experimental platform for the edge-lit backlight architecture was also established in collaboration between B&O, and DTU Fotonik and with support from TekPartner (Danish based SMV) transferred this to the platform. This included a playback mode, useful for visual testing. Thus having the capability of displaying and visually assessing the results of the backlight algorithms from early in the project has been an important aspect and asset to the work.


The backlight display model and optimization algorithms have been developed in five steps: 1) A basic physical and mathematical model and well-defined optimization set-up focussed on mse (mean square error) distortion was presented. 2) An extension to include the energy consumption in the optimization cost function and expressing the distortion in a perceptual domain, correlating much better with visual impression than luminance mse distortion. This optimization was based on gradient descent (GD). The achievements of Steps 1 and 2 provided an optimal solution for image by image processing. 3) A fast version of GD, called block-based GD (BBGD), was developed. This also included quality considerations for the temporal dimension of video. 4) For video sequences a power-control over time was introduced based on careful analysis of the image by image trade-off of quality and power. 5) Measurements of the ambient light was included, the influence on visual quality evaluated and included in the adaptive backlight control. Thus optimal backlight evaluating perceptual quality and energy consumption was introduced at the image level and extended to the temporal dimension to include video fast algorithms and quality and power control.   


The optimization-based backlight dimming algorithms derived and developed provide optimization of visual quality (contrast) and energy consumption in backlit displays as well as a weighting between energy consumption and quality. The figure displays quality (measured by Peak Signal-to-Noise Ratio, in dB) vs. power consumption normalized by that of full backlight. The proposed algorithm allows setting a relative weight (q) on power consumption. Thus a whole range of solutions are available whereas algorithms in the literature only provide one solutions (single points in the figure), or at best some parameters, which are not clearly defined in terms of neither quality nor energy consumption. It is seen that the proposed algorithm can provide better quality at fixed power level or achieving a fixed quality at a lower power level. This frame work provides an excellent basis both for academic analysis as well as further developing efficient backlight dimming algorithms.


As the goal of the image processing is to improve visual quality, a number of subjective tests have been performed. The assumptions and quality evaluations were tested and evaluated. E.g. the correlation of the modelling of display quality was reported and correlation of subjective quality and objective metrics were evaluated finding high correlation for selected objective metrics for images.


Quality (measured by Peak Signal-to-Noise Ratio, in dB) vs. power consumption normalized by that of full backlight, compared with methods from the literature.


A technical result of the research was demonstration and test of energy efficient control of LED backlight display for HDTV and UHD including measurements of ambient light. The primary set-up spanned the combination of a range of the display (peak white) brightness values of 75, 200, 490 cd/m2 (the latter being the max. value for the given TV) and a range of ambient light level: 0 (lights off), 5 lux and 60 lux and it was based on the experimental platform of a Full HD edge-lit display. This was used for visual testing Bang & Olufsen and also as the primary demonstration of the project. This test and demo was based on initial pilot testing only changing the levels of ambient light, modeling of the contrast and LCD leakage, and the visual impact of backlight dimming. The results showed significant visual improvement of quality of a local backlight dimming providing a 30% reduction of display power consumption at high quality. It showed that there is a significant interaction of display peak with and ambient light and a balance of ambient light and peak white was preferred. This leads to the suggestion for a low power or home cinema mode in low ambient light, reducing display power consumption by as much as a factor of 10 overall, i.e. to 10%. Reducing artificial ambient light, e.g. living room light in the evenings as part of this set-up will also reduce household energy consumption.


3. Energy savings


The topic is as mentioned LED backlight dimming in TV displays. The energy consumption of the display is directly proportional to the LED values, thus dimming the LED backlight directly translates into energy savings. The display constitutes the major part of the energy consumption. Optimal backlight dimming has been derived and demonstrated. This can provide energy savings and higher contrast on LCD screens with segmented LED backlight.


In the initial experiment on relatively dark images, results depicted in the figure above display the quality (by Peak Signal-to-Noise Ratio) vs. normalized power consumption with full backlight consuming 1. The backlight energy consumption is at high quality reduced up to 58% compared with a state-of-the-art dimming algorithm (Albrecht comparing at the same PSNR quality level) in a given test, and reduced to only 30% compared with full backlight. For a representative selection of the ISO/IEC standard test sequences, dominated by bright TV sequences, a saving of 20-25% compared with Albrecht was achieved and the high quality at a PSNR of 45 dB was achieved at about 50 % of full backlight. At highest quality the saving was about 30% compared to full backlight, at the same time as providing significantly better quality when tested on challenging dark high contrast sequences. As the backlight consumes most of the energy this shows the potentials for energy savings for high quality display.


Inspired by and based on the project results, Bang & Olufsen has developed and implemented a 1.5D dimming algorithm for a 2x8 segment LED dual edge-lit display. This novel backlight dimming solution has been launched in the new BeoVision Avant TV product family developed and manufactured by Bang & Olufsen.

This saving is achieved at the same time as the quality is enhanced. Higher energy savings are possible albeit at the expense of graceful degradation of image quality. More work is required to achieve the full benefits of quality-energy optimization.  


In this research other important results have been produced that have proven essential to optimising the power-quality trade-off for practical industrial TV applications with local backlight dimming of the LCD panel. These include the visibility of temporal effects in the backlight dimming algorithms, leading to important constraints to be put on the frame-to-frame changes in backlight drive. Another outcome is knowledge about the influence of the viewing angle on the amount of backlight leakage and hence, the optimisation of the algorithm with viewing angle as an independent variable. Furthermore the influence of the ambient light on the perceived quality has been studied. Finally, perception methods for the evaluation of backlight dimming algorithms have been developed, combined with a wide set of video sequences specifically suitable for the assessment of dimming algorithm side-effects.


In the latest Ultra HD (UHD) TV family from Bang & Olufsen, BeoVision Avant, local backlight dimming is applied using two different dimming technologies as illuastrated above. For the larger 85 and 75 inch screens direct-lit 2D dimming is used. In the 55 inch BeoVision Avant edge-lit dimming is used with 160 LEDs organized on the left and right edge in 2x8 segments having 10 LEDs each.




Selected papers



Søren Forchhammer
Professor, Group Leader
DTU Fotonik
+4545 25 36 22