Possibly, the most fundamental general statement is that the overall digital image representation must be as close as possible to the ideal original continuous image, as seen by the observer’s eye. This was the underlying basic concept to formulate the specific design principles that will be discussed below.
2.1. Minimal Colour Fringe Error
One principle is the requirement of minimal CFE, when using sub-pixel rendering, together with good colour rendering and without visible chromatic error at the edge of spatial patterns. Colour fringe error is most disturbing in the achromatic case, at the border of black characters over white background. RGB-stripe architectures fail to avoid CFE, especially without spatial filtering (blurring). In Figure II/5, it can be observed that a “non-rgb” sequence (e.g. g-b-r or b-r-g) will not be perceived as white on dark background but rather as two neighbouring colour pixels (cyan-red for g-b-r and blue-yellow for b-r-g).
ClearType technology was developed to prevent this in text rendering70 but it enables only the horizontal smoothing of letters or similar shapes. To facilitate the perfect blending of the single-colour light from the sub-pixels, the so-called “checkerboard” arrangement of the sub-pixels seems to yield good results. This means that the sub-pixels are positioned in an altering way so that no neighbouring of the same primary colours can arise. This kind of sub-pixel arrangement reduces the colour fringe error since better colour blending is possible, contrary to the RGB stripe or other similar architectures where the identical sub-pixels are positioned in the neighbourhood of each other.
2.2. Modulation Transfer Function
High modulation transfer function (MTF) limit stands for good display resolution and it is originally used to describe the “contrast transmission” capabilities of optical devices (e.g.
lens systems in photography)74 but the definition can be easily adopted for display devices since it is a curve representing the ratio of the output and input contrast as a function of line frequency: MTF(ν)=Mo/Mi where Mo and Mi indicate the modulation (or contrast) of the output and input images of square wave grating (altering black and white lines) of line frequency ν. Modulation M is defined as M=(Lmax-Lmin)/( Lmax+Lmin) where Lmax is the maximum and Lmin is the minimum luminance of the grating. The simplified definition of the MTF to display devices is the maximum number of representable altering vertical black and white line pairs over a given area. It is essential for a display to be able to present the
largest possible number of pairs over a given area. In the conventional RGB-stripe sub-pixel architecture, it is impossible to generate very thin lines (e.g. 1 or 2 sub-sub-pixels width) of any otherwise displayable colour including white or grey. The minimum width of such lines is the width of the whole pixel of the architecture.
2.3. Isotropy
Beside the possibility of displaying thin white or grey lines, a more severe prerequisite is the chance of displaying separate dots of any colour. If the modulation transfer function is high for both the vertical and the horizontal directions, this requirement can be fulfilled.
This consideration leads us to the next important principle, namely isotropy, which is the directional independence of the sub-pixel pixel architecture. For the conventional RGB-stripe sub-pixel architecture, simple horizontal and vertical line rendering can be accomplished easily but sub-pixel rendering will not work in vertical directions. (Note that several FPDs on the market offer 90° rotated screen built to achieve letter dimension).
Anyway, tilt lines and calligraphic letters cannot be displayed optimally in the conventional RGB-stripe sub-pixel architecture.
A synopsis of the first three principles (2.1-2.3) may lead to the derivation of another design principle: It should be possible to draw thin and long rectangular stripes not only vertically and horizontally, but in any direction, and, within these thin stripes, the summation of the colour of the crossed sub-pixels should be able to yield any colour within the colour gamut, including grey and white.
Though, from the ophthalmologic field it is well-known that the contrast sensitivity of the HVS is poorer for oblique than for vertical and horizontal orientations for the case of middle and high spatial frequencies, which is known as the ‘oblique effect’75, directional independence of a pixel structure in this sense grounds for the more potential elimination of CFE. The colour fringe error emerging from the not complete summation of the colour of the sub-pixels is visually multiplied if the neighbouring same primary sub-pixels merge and create single colour blocks (as seen at the legs of letter ‘m’ in Figure II/5). CFE is thought to be more robust and the oblique effect is only of minor importance to think about (certainly, it is not a disadvantage if the display outperforms the HVS).
2.4. Luminance Resolution
In the human visual system, the relatively low spatial resolution of short-wavelength sensitive receptors (S-cones) in the retina3 would allow the placing of blue (or bluish)
sub-pixels in a sparser manner (see the schematic representation in Figure I/1). In order to reach the proper colour balance (i.e. the proper display white point) in this case, the area of these mostly S-cone stimulating bluish sub-pixels should be larger than the area of other sub-pixels. This was obtained for the PenTile Matrix structures.73 But blue pixels cannot exhibit high luminance and therefore, their larger size would visually divide the architecture and block the visually uniform blending of the light output from the sub-pixels. Thus, these large-sized blue sub-pixels would evoke visible textures in those areas where the sub-pixel architecture is intended to look homogenous. This might be the reason why a vertical white line built up from a non-RGB sequence look like two neighbouring coloured line pairs and why the disturbing colour fringes are visual at the boundary of the letters when no blurring is applied (Figure II/5).
To overcome this, instead of less large blue sub-pixel, using more blue sub-pixels of normal size (same size and number as the other sub-pixels) and (e.g. in couples) of the same address seems better, allowing for the “smaller blue resolution” criterion. It is well known that in HVS, it is the luminance channel that carries the finest spatial detail in the image. It was shown in several papers that the discrimination and appearance of small field colours (see Section 1.5.1 of Chapter I) or fine-detailed patterns differ from that of the medium sized ones.76-79 The luminance of the green or greenish sub-pixels is highest and therefore, referring to the above considerations, green resolution may be greater than the resolution of the other display primaries. In the three-colour primary Bayer pattern80 arrangement used recently in the CCD filtering mosaic array of digital cameras, green resolution is twofold (a checkerboard arrangement with the red-green-blue ratio equal to 1:2:1). But, again, it must be kept in mind that, since displays are output devices, the appropriate white point is a very important criterion.
2.5. High Aperture Ratio
The visually useless area of the display, i.e. the area used for wiring or other electronic components where there is no emitted light, should be minimized to achieve good luminance efficiency. To achieve this, we must think about how to cover the plane completely with simple regular two-dimensional figures that are easy to obtain. As it is well known from theoretical considerations, we can use only a limited set of such regular shapes: triangles, rectangles and hexagons. The best spatial arrangement for sub-pixel image rendering would probably be a stochastic pattern in which the human visual system cannot discover any systematic order, similar to ink-jet printing. But such a pattern would
probably imply undesirable manufacturing costs, not mentioning that this arrangement would work only at relatively higher resolution, which contradicts the basic idea of sub-pixel rendering.
After revealing the sub-pixel rendering possibilities of the modern display devices together with the benefits of an enhanced colour set, namely multi-primary technology, the major message of the chapter will be reflected on. What happens if these two features are both present? It will be shown that it is possible to examine multi-primary imaging and sub-pixel rendering simultaneously.81 New architectures will be introduced as possible arrangements for more than three primaries and compared to the RGB-stripe from several viewpoints. Finally a colour rendering routine will be shown that exploits the nature of our visual system placing sub-pixel rendering on the basis of colorimetry.