Introduction
1.1 Categorization of Display Artifacts
Most display artifacts are related to not only the display stimuli, but also the perception of human vision system (HVS). On the stimulus side, the first-order parameters include luminance, chromaticity, temporal frequency (field rate), spatial frequency (grating), and moving velocity. On the perception side, since HVS is three-dimensional, the detection thresholds are different between the luminance and chromaticity domain. Table 1 enumerates the common artifacts detected in different conditions.
Table 1: Categorization of Display Artifacts
Detected in Luminance Detected in Chromaticity Spatial color
Aliasing Subpixel dithering
Sequential color
Still target Luminance flicker Chromatic flicker Sequential color
Moving target Motion blur Color breakup
The first row represents scenarios like inspecting a white screen on a conventional color spatial LCD. The mura is detected if any luminance difference is perceived at different locations. Color shift (e.g. due to viewing angle difference) may still be perceived even when mura is absent from perpendicular measurement because HVS has higher sensitivity to chromatic difference at low spatial frequency. The contrast sensitivity functions (CSF) of still
2
target in the luminance and chromaticity domain had been well established. In both temporal and spatial domains, luminance CSF is a band-pass filter while chromaticity CSF is low-pass.
The second row speaks for the same pattern on a field sequential display. If the field rate is lower than the critical flicker fusion frequency threshold, luminance flicker may be perceived. Note that the RGB primaries have different luminance so luminance flicker is always detected before chromatic flicker. In other words, isolating chromatic threshold from luminance on a field sequential display is challenging.
The third row brings up the spatial resolution issues on conventional LCDs. The aliasing artifact is only perceivable when the human eye can resolve the pattern at pixel level.
Recall that luminance CSF is more sensitive than the chromaticity CSF at medium spatial frequencies. The principle of subpixel dithering is to supplement luminance information without chromaticity being detected.
The last row introduces movement of the target. In this case, the gaze position of the observer determines how the artifacts are perceived. The stimulus is called stable if the target and gaze position are in sync, i.e., the target is perfectly pursued by the eye movement.
Otherwise, the stimulus is called unstable. Unstable stimulus can be caused by different types of eye movement – fixation, smooth pursuit, and saccade. Notice that the HSV has very different sensitivity in these three movements. Overlooking this fact and use the vision models for fixation to predict artifacts in the other two is a common mistake in CBU-related studies.
3
1.2 Field Sequential Display
The field sequential display synthesizes colors in the time domain. By quickly flashing the red, green, and blue field one after the other, the observer is unable to distinguish the time difference between the three channels in Figure 1.
Figure 1: Field sequential display separates RGB fields.
The field sequential technology has been successfully used in TI’s DLP-based projectors, which use fast-switching micro mirrors to produce gray-levels and a color wheel to produce the primaries as shown in Figure 2. In this way, resources can be shared by the three channels and hardware cost can be greatly reduced.
If such technology can be adapted for the LCD (i.e., FSC-LCD), not only its luminance efficiency can be increased to 3X, which equates to considerable power savings, but also the hardware cost can be cut down to 80% because the costly color filter process on the glass substrate can be eliminated. Unfortunately, unlike DLP, the slow response time of liquid crystals limits the highest frame rate of field sequential LCD and results in severe color breakup artifacts. Due to its nature of synthesizing colors temporally, FSC-LCD suffers from the color breakup phenomenon when eye movement and stimulus movement are out of sync.
When the red, green, and blue components of the same stimulus project onto different locations of observer’s retina, color breakup is bound to happen depending on stimulus and
4
viewing conditions.
Figure 2: How DLP projection works and Digital Micro-mirror Device (DMD) chip.
Nevertheless, the field sequential LCD has the advantage of flexible backlighting schemes. The LED backlights are capable of generating arbitrary waveform for each of the three primaries or their combinations in arbitrary order [1].
1.3 Color Breakup
The most infamous artifact on field sequential displays, the color breakup phenomenon (CBU) shown in Figure 3, occurs when the red, green, and blue components of the same object project onto different locations of retina upon eye movement. Originated in the 80s, the study of field sequential display revived in the recent years for the temptation of high optical efficiency, high spatial resolution, and low manufacturing cost to the liquid crystal display technology (LCD) [2][3]. Suffering from slow response time, LCD is prone to the CBU artifacts. Despite its long research history, the foundation of CBU is still difficult to analyze due to the tangling causing factors such as the target movement, eye movement, field rate, target luminance, target pattern, primary colors/waveforms, ambient light, eccentric angle, viewing conditions, etc. [4].
5
Figure 3 : Motion-induced the color breakup in field sequential display.
1.4 Organization
In this thesis, three approaches are proposed for minimizing display artifacts which are induced by eye movement. From slow to fast, these artifacts are: viewing angle-dependent color shift caused by head movement, color breakup caused by smoothly pursued eye movement, and color breakup caused by saccadic eye movement. A platform for evaluating perceivable color breakup of human eye is proposed in Chapter 2. In Chapter 3, for saccadic color breakup, a custom-made electro-oculogram circuit is used to detect the events of saccadic eye movement. In Chapter 4, for pursued color breakup, an eye-tracker is used to detect the gaze velocity such that the image chroma can be reduced for suppressing color breakup at run-time. In Chapter 5, for viewing angle-dependent color shift, an infrared sensing mechanism is used to detect the head position. The LCD panel transmittance and backlights are modulated to compensate for the color shift accordingly. Finally the conclusions and the future works are given.
6