Flat panel display signal processing

Televisions (TVs) have shown considerable technological progress since their introduction almost a century ago. Starting out as small, dim and monochrome screens in wooden cabinets, TVs have evolved to large, bright and colorful displays in plastic boxes. It took until the turn of the century, however, for the TV to become like a ‘picture on the wall’. This happened when the bulky Cathode Ray Tube (CRT) was replaced with thin and light-weight Flat Panel Displays (FPDs), such as Liquid Crystal Displays (LCDs) or Plasma Display Panels (PDPs). However, the TV system and transmission formats are still strongly coupled to the CRT technology, whereas FPDs use very different principles to convert the electronic video signal to visible images. These differences result in image artifacts that the CRT never had, but at the same time provide opportunities to improve FPD image quality beyond that of the CRT. This thesis presents an analysis of the properties of flat panel displays, their relation to image quality, and video signal processing algorithms to improve the quality of the displayed images. To analyze different types of displays, the display signal chain is described using basic principles common to all displays. The main function of a display is to create visible images (light) from an electronic signal (video), requiring display chain functions like opto-electronic effect, spatial and temporal addressing and reconstruction, and color synthesis. The properties of these functions are used to describe CRT, LCDs, and PDPs, showing that these displays perform the same functions, using different implementations. These differences have a number of consequences, that are further investigated in this thesis. Spatial and temporal aspects, corresponding to ‘static’ and ‘dynamic’ resolution respectively, are covered in detail. Moreover, video signal processing is an essential part of the display signal chain for FPDs, because the display format will in general no longer match the source format. In this thesis, it is investigated how specific FPD properties, especially related to spatial and temporal addressing and reconstruction, affect the video signal processing chain. A model of the display signal chain is presented, and applied to analyze FPD spatial properties in relation to static resolution. In particular, the effect of the color subpixels, that enable color image reproduction in FPDs, is analyzed. The perceived display resolution is strongly influenced by the color subpixel arrangement. When taken into account in the signal chain, this improves the perceived resolution on FPDs, which clearly outperform CRTs in this respect. The cause and effect of this improvement, also for alternative subpixel arrangements, is studied using the display signal model. However, the resolution increase cannot be achieved without video processing. This processing is efficiently combined with image scaling, which is always required in the FPD display signal chain, resulting in an algorithm called ‘subpixel image scaling’. A comparison of the effects of subpixel scaling on several subpixel arrangements shows that the largest increase in perceived resolution is found for two-dimensional subpixel arrangements. FPDs outperform CRTs with respect to static resolution, but not with respect to ‘dynamic resolution’, i.e. the perceived resolution of moving images. Life-like reproduction of moving images is an important requirement for a TV display, but the temporal properties of FPDs cause artifacts in moving images (‘motion artifacts’), that are not found in CRTs. A model of the temporal aspects of the display signal chain is used to analyze dynamic resolution and motion artifacts on several display types, in particular LCD and PDP. Furthermore, video signal processing algorithms are developed that can reduce motion artifacts and increase the dynamic resolution. The occurrence of motion artifacts is explained by the fact that the human visual system tracks moving objects. This converts temporal effects on the display into perceived spatial effects, that can appear in very different ways. The analysis shows how addressing mismatches in the chain cause motion-dependent misalignment of image data, e.g. resulting in the ‘dynamic false contour’ artifact in PDPs. Also, non-ideal temporal reconstruction results in ‘motion blur’, i.e. a loss of sharpness of moving images, which is typical for LCDs. The relation between motion blur, dynamic resolution, and temporal properties of LCDs is analyzed using the display signal model in the temporal (frequency) domain. The concepts of temporal aperture, motion aperture and temporal display bandwidth are introduced, which enable characterization of motion blur in a simple and direct way. This is applied to compare several motion blur reduction methods, based on modified display design and driving. This thesis further describes the development of several video processing algorithms that can reduce motion artifacts. It is shown that the motion of objects in the image plays an essential role in these algorithms, i.e. they require motion estimation and compensation techniques. In LCDs, video processing for motion artifact reduction involves a compensation for the temporal reconstruction characteristics of the display, leading to the ‘motion compensated inverse filtering’ algorithm. The display chain model is used to analyze this algorithm, and several methods to increase its performance are presented. In PDPs, motion artifact reduction can be achieved with ‘motion compensated subfield generation’, for which an advanced algorithm is presented.