Making Large-Sized TFT-LCD TVs Better

TFT-LCDs have made rapid and significant advances, becoming the technology of choice for large-screen HDTV applications. But more advanced color simulation tools and image-processing algorithms are making even more improvements possible.

by C. T. Liu, A. Wang, H. J. Hong, Y. J. Hsieh, M.S. Lai, A. Tsai, T. M. Wang, M. J. Jou, W. C. Chang, S. L. Sui, J. H. Liao, and M. F. Tien

AS TFT-LCD TECHNOLOGY becomes the mainstream technology that will replace CRTs in TV applications, manufacturers are continually looking for ways to improve color reproduction and image quality. TFT-LCD systems are more complicated to optimize than CRTs, but at the same time they offer optimization possibilities far beyond those of existing CRT displays.

Because TFT-LCD color generation is a result of the interaction of several components – the emissive spectrum of the light source, the transmissive spectra of the LCD panel, the polarizers, and other optical films – we looked at some advanced color-simulation techniques that take into account all the components along the optical path of this system. Digital calibration and new image-processing algorithms offer additional opportunities for LCD image enhancement. We will describe a few of the possibilities, but certainly there are others.

Color Simulation

TFT-LCDs with a wide color gamut of 76% NTSC have been produced. But for professional monitor and TV applications, the color saturation must be increased to beyond that required for a 76% color gamut. We have developed a simulation tool that optimizes the emissive spectrum of the light source and the transmissive spectra of all the optical components, including the optical films, polarizers, color filters, and liquid crystal. Dark-state color is also an important attribute for LCD TVs, and this simulation tool is capable of predicting the color gamut and dark-state color for LCD panels.

 

fig_1_tif

Fig. 1: The red, blue, green, and white (RGBW) charts show plots generated by a conventional calibration method, taking into consideration only the color filter and the light source (open squares), versus an improved calibration technique that takes into account the liquid-crystal characteristics, the cell gaps in the panels, the polarizer, and all other optical films. The improved simulation (open triangles) shows good consistency with the actual color gamut and white color chromaticity (black dots). Note that the x-axis value varies by 0.01 for R, G, and B, while it varies by 0.03 for W.

 

Conventional color simulation only takes into consideration the color filter and the light source, and therefore predicts that the panel will exhibit colors that are quite different from those that actually appear (Fig. 1, open squares). By taking into account the liquid-crystal characteristics, the cell gaps in the panels, the polarizers, and all other optical films, the improved simulation shows good consistency in comparison with real products (Fig. 1, open triangles). By using this simulation method, both color gamut and white color chromaticity can be predicted with accuracy, and the difference in the NTSC ratio between the simulated results and the real products is only about 3%. The maximum discrepancy in chromaticity is about 0.01.

In this simulation, the most challenging task is the optimization of the transmission spectrum of the liquid crystal itself because it is affected by the pixel voltage, operational temperature, and details of the pixel design. The improvement is most significant for white, in which the x-axis value varies by only 0.01 for R, G, and B, and by 0.03 for white.

By applying our simulation tool, we have achieved a color gamut of 96% NTSC for an LCD using a CCFL light source and a color gamut of 109% NTSC for an LCD using an LED light source. When an LED backlight is used in an LCD that is not optimized, R and G are greatly improved, but B is actually not as good compared with an LCD that uses a conventional CCFL backlight. After optimization, B also performs better even when a conventional CCFL backlight is used.

 

fig_2_tif

Fig. 2: This chart plots the required gray levels using alternative calibration techniques. If the dark grays are adjusted to the desired color temperature, the contrast ratio will be sacrificed, and only gray levels greater than 200 can fall within the u'v' < 0.01 circle for a target temperature of 10,000K (orange curve). By using the new calibration approach, all gray levels greater than 24 can fall within that circle (pink curve).

 

fig_3_tif

Fig. 3: Contrast enhancement is generally implemented by expanding the gray-level distribution in a process called histogram equalization, but this approach may create unnatural-looking colors. By incorporating edge information in the calculation, the adjusted image looks natural and the resulting histogram retains the same general shape as the original, even as more bright pixels appear in the image, i.e., even as the contrast has been increased. To observe this, look at the original image with its corresponding histogram (left) and compare it to the contrast-enhanced image with its corresponding histogram (right).

 

Color-Temperature Calibration

The goal of color-temperature calibration is to adjust the neutral gray level so that it maintains the target color temperature from white to black. For the conventional calibration process, color temperature can be controlled only above a certain gray level. Below that, the color temperature increases much higher than expected. If we adjust the dark grays to the desired color temperatures, the contrast ratio will be sacrificed. We have investigated several possible approaches to limit the uncontrolled gray levels while keeping the contrast ratio constant. For the existing approach, only gray levels greater than 200 can be within the u′v′ < 0.01 circle of the target temperature of 10,000K (Fig.2, orange curve). By using the new calibration approach, all gray levels greater than 24 will fall within that circle (pink curve).

In performing this process, it is very important to maintain a smooth transistion from gray level 0 to 255 while adjusting the color temperature and gamma. It is particularly important to maintain smooth gamma curves for R, G, B, and W. In addition, we have tried to find an automatic calibration method instead of being forced to have an experienced engineer set the parameters manually. Based on both the data and perceived images, this calibration method can provide very good results. However, a greenish or yellowish tinge can still be observed from an off-angle position. More work still needs to be done in this area to improve off-angle display quality.

Contrast Enhancement

The main objective of contrast enhancement is to expand the gray-level distribution in order to make gray-level utilization more efficient. Histogram equalization is usually adopted to solve this problem by adjusting the entire image frame according to the specifications of the frame regardless of the spatial information. But such an approach may generate results with an unnatural look, otherwise known as artifacts.

 

fig_4_tif AU Optronics Corp.

Fig. 4: In this figure, the top images use fixed dithering patterns with some unexpected streaks appearing in both the horizontal and the vertical directions. These cascading effects are caused by the subsequent fixed-frame-rate control (FRC) patterns of upstream video. The lower images use dynamic dithering patterns (DFRC) that overcome the problem.

 

fig_5_tif AU Optronics Corp.

Fig. 5: A 6-bit image of color bars and a blue sky is displayed on a TFT-LCD panel. Without the application of dynamic frame-rate control (DFRC), the 6-bit color bars consist of 64 sequential gray levels, with distinguishable vertical stripes (left). After applying DFRC, the color bars exhibit 256 gray levels and the stripes disappear. Similarly, the 6-bit rendering of a natural image of a blue sky that contains some irregular contours (left) can also can be dithered by DFRC, resulting in a dramatically improved image (right).

 

Our proposed solution takes pixel locality into consideration in order to avoid the generation of images with unreasonable color levels. This novel technology enhances image contrast according to edge information. A dedicated detector is used to extract the edge-position information, and the contrast-enhancement ratio is based on this parameter. The pixel intensity increases or decreases, depending on the pixel distribution across the edge. By using the edge information, the adjusted image will look more natural [see Fig. 3, upper left, and compare it to the contrast-enhanced image with its corresponding histogram (Fig. 3, upper right)]. For the proposed method, the histogram's shape is kept the same as that for the original image (Fig. 3, lower left), while the contrast has been enhanced (Fig. 3, lower right).

By using a new technique, a dynamic gamma was developed to modify the gamma curves according to a weighting factor that considers the gamma curves of previous frames. The gamma curve of each frame is created by the new histogram information generated by both the proposed contrast-enhancement algorithm and the parameters from previous frames. This approach ensures a smoothly changing sequence for both the colors and brightness of consecutive frames. Thus, users will not notice a dramatic or discontinuous change in the colors or brightness.

Dithering

Image dithering is the process of creating smooth continuous tones from spatial patterns of monochromatic dots. This is normally accomplished using frame-rate control (FRC) and is implemented by varying the tones of pixels in temporally consecutive frames.

Conventional FRC utilizes a fixed pattern to perform the image dithering, but occasionally the fixed dithering pattern may produce unwanted textures generated from subsequent FRC patterns of upstream video.

This FRC cascade effect creates abnormal textures that degrade the display quality to unacceptable levels. An advanced FRC algorithm with dynamic patterns, referred to as Dynamic FRC (DFRC), has been developed at AU Optronics Corp., and it greatly improves the conflict of cascaded FRC (Fig. 4).

To compare the results of this approach before and after the application of DFRC, a 6-bit image of color bars and a blue sky was displayed on a TFT-LCD panel (Fig. 5). The "before" 6-bit color bars consist of 64 sequential gray levels, with distinguishable vertical stripes (left). By applying DFRC, the color bars exhibit 256 gray levels and the stripes disappear. Similarly, in the same figure, the 6-bit rendering of a natural image of blue sky that contains some irregular contours (left) can also can be dithered by DFRC, resulting in a dramatically improved image (right).

Conclusion

The color and image performance of large-sized TFT-LCD TVs is in an evolving state. We have introduced several new techniques that begin to address the scope of possible improvements. We can surely expect that even more advanced techniques will continue to be developed in the areas of digital processing and component improvement by independent suppliers, LCD-panel makers, and TV-system makers. •

 


C. T. Liu is Vice President in charge of the new AUO Technology Center at AU Optronics Corp., No. 1, Li-Hsin Rd. 2, Science-Based Industrial Park, Hsinchu 300, Taiwan, R.O.C.; telephone +886-3-500-8800 x7222, fax +886-3-5679074, e-mail: ctliu@auo.com.