The invention concerns power management and image enhancement in visual display devices and, more particularly, in liquid-crystal display devices.
Visual display devices are ubiquitous in battery-powered portable electronic devices such as notebook computers and mobile, hand-held telephones where, typically, they are the largest consumers of battery power. For example, in mobile devices equipped with thin-film transistor (TFT) liquid-crystal displays (LCD) utilizing backlight illumination, the LCD panel consumes more than 30% of the device power and the backlight typically consumes more than 75% of the LCD power. Thus, for conserving battery power, there is primary interest in minimizing the power consumption of the display device.
An LCD screen typically includes an array of liquid-crystal pixels arranged as a plurality of rows each having a plurality of pixels, arranged in columns, with each pixel capable of displaying any one of a plurality of luminance values of a gray scale and the corresponding chrominance values. Each pixel has its own liquid crystal cell, a dedicated thin-film transistor, and a dedicated capacitor. The electrical field of the capacitor controls the orientation of the liquid crystals within the cell, determining the optical transmissivity of the cell and thus its luminance when lit by a backlight. The capacitor is charged and discharged via its transistor. Device activation typically is row-by-row, so that, at any one time, all column lines are connected to a single row.
For saving power in an LCD device, dynamic backlight control can be used in playing back a video, a movie clip, or any other form of multimedia data such as still images, gaming or animation content. Frame by frame in dynamic backlight control, backlight brightness can be scaled down to a value that is just enough to display each video frame with sufficient quality, while simultaneously transforming the frame so as to compensate for the change in backlight brightness.
We have recognized that on applying dynamic backlight scaling to images of a video scene or movie, video quality may be impaired by perceived flicker. As a particular cause of flicker we have identified backlight changes within a scene, and we have recognized that flicker can be ameliorated when video backlight brightness is adjusted scene by scene, for the frames of each video scene to be displayed with essentially the same backlight brightness.
For dynamically controlling backlight brightness scene-by-scene, video scene-change information may be available explicitly beforehand. Alternatively, scene changes can be detected in real time under attendant constraints, e.g. those imposed by limited processing power of a client device such as a cell phone or a personal media player.
Video scene changes can be determined prior to their ultimate use in a client device, with such prior determination being made by the client device, by a server computer where the video is stored, or by any other suitable processor capability. From a server computer, available scene-change information can be downloaded along with the video to the client device where the video will be displayed. Alternatively, scene-change information as determined on the client device or elsewhere can be stored on the client device for later use during actual video display.
As an example of scene-based backlight scaling, in
As an example of backlight scaling, in
Determining the Backlight for a Scene
Given the scene boundaries, pixel value statistics can be determined for each frame within the scene. Based on these statistics, a backlight can be chosen such that almost all the pixels within the scene can be faithfully rendered. For example, in one embodiment, the maximum among all frames within the scene is chosen, of the 97th percentile of the pixel values for each frame. Then the backlight is set so as to represent this 97th percentile faithfully. To this end, knowledge is required of the perceived brightness under the default backlight for that 97th percentile pixel. For example, if the 97th percentile is 150 and has a corresponding brightness of 100 nits, and the maximum possible pixel value of 255 has a brightness of 300 nits, then the backlight can be scaled down by a factor of 100 nits/300 nits≈0.33. More generally, if, based on the pixel statistics, the maximum pixel value to be displayed faithfully is p and has a corresponding brightness of max_b2, and if the maximum possible brightness of the display for a pixel value of 255 is max_b1, then the backlight can be scaled down by a factor of max_b2/max_b1.
The pixel-value to pixel-brightness mapping is usually not a linear relationship. It is typically defined by an exponential or “gamma” curve:
where b is the brightness for pixel value x, max_b is the maximum pixel brightness for a pixel value of 255, and γ (gamma) is the parameter that controls the shape of the curve. The gamma values for each display can be derived by measuring brightness for several pixel values, and fitting the gamma curve given in Equation 1 to the measured data.
Determining the Pixel Transform for a Frame
As with the backlight, the display gamma curve can be used for determining the pixel transform. The transform can be derived as
x′=min(m·x, 255), (2)
using the transform scale factor
where max_b1 denotes the maximum brightness for the backlight level used by the display at its default backlight setting, max_b2 denotes the maximum brightness for the target backlight, and γ denotes the measured gamma value for the display.
The gamma values for the display can be different for each of the primary colors, red (R), green (G), and blue (B). As it is important to display all colors faithfully, the methods for determining the backlight and the transforms are modified as follows:
1. For the backlight, determine the target backlight value by using pixel-value statistics for each of R, G, and B, and then choose the maximum of the three target backlights. Choosing the maximum ensures that all colors in the scene can be rendered faithfully.
2. For the pixel transform, use Equation 2 for each of the colors, R, G, and B separately. The transform scale factor for each color is mR, mG, and mB, respectively.
3. For a different color space, such as the YUV space (where Y denotes brightness, and U and V represent color), there is a known invertible linear correspondence to the RGB color space. This transformation can be written as
where M is an invertible 3 by 3 matrix. Then the transformed values of Y, U, and V are given by
If the matrix T is such that the diagonal elements dominate, then, for reducing computational cost, T can be approximated by the diagonal matrix formed from T by zeroing out its off-diagonal elements, so that each of the terms, Y, U and V simply is scaled by a respective scale factor.
A statistic of pixel component values can be determined in a preliminary scan of a sequence of given frames, for example, or in the course of a prior application of the method or of one of its parts. In determining a backlight brightness value and/or a respective transform, display characteristics of a target display device can be taken into account. Such display characteristics of the display device can include pixel-value-to-brightness characteristics at different backlight values for color components and/or luminance component, and/or brightness-to-power characteristics of a color component and/or luminance component. Where demarcations of visual scenes are to be determined, they can be obtained in a preliminary scan of a sequence of frames, e.g. in the course of a prior application of the method or of one of its parts.
A transform can be determined for a quality of visual perception of the transformed frames displayed on the display device with the selected backlight value. For example, the determination can aim at substantial matching of brightness of a transformed pixel at the selected backlight value with brightness of a given pixel at a given brightness value, and visual perception can depend on visual contrast, color hue and color saturation. Conveniently, transform information can be stored in a lookup table. A display device can be disposed in a consumer device such as a cell phone, personal media player or large-screen TV.
Techniques of the invention can be implemented in distributed fashion, with functions carried out by modules in operational communication for cooperation in effecting an ultimate display. Such cooperation can be facilitated by metadata of scene demarcations and pixel statistics, for example, with the metadata accompanying the pixel data or being stored in a separate file. The techniques can be implemented in hardware, firmware or software.
Calibrating the Display Device
As different display devices such as LCD display panels have different luminance and/or chrominance display characteristics, it is advantageous to calibrate a device for optimal performance. For calibrating an LCD device, the following method can be used:
1. Measure the power consumed by the backlight at various levels of backlight brightness from minimum to maximum brightness, thus establishing a relation P=p(b) between backlight power, P, and backlight brightness, b.
2. Determine the maximum backlight brightness, bmax, at which the power saved by backlight reduction is greater than the power consumed for transforming the video frames. This power, for transforming the video frames, may be expended in a separate subsystem such as a microporcessor connected to a display subsystem. For a given number of levels of backlight brightness, typical video sequence brightness statistics can be used together with the relation P=p(b) to determine the actual brightness values between 0 and bmax so that average power is minimized for a typical video sequence. The backlight brightness levels may be evenly spaced, as is indicated when the levels all are equally likely. Or, if the typical brightness distribution has a strong mode around a particular value, then more brightness values may be used to advantage near that mode.
3. For each of the color components, measure the brightness for various pixel values, and fit a gamma curve to the measurements to determine the value of gamma for each color component. Now the pixel transformation can be determined as described above, for any target backlight value.