Measuring Pursued Color Breakup

Due to its nature of synthesizing colors temporally, the field-sequential-color liquid-crystal-display (FSC-LCD) suffers from 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 the observer’s retina, color breakup is bound to happen depending on stimulus and viewing conditions. Therefore, a robust model of predicting color breakup is demanded by designers of field sequential displays. To derive such a model, statistical data must be collected from subjective experiments with human subjects. However, color breakup is a spontaneous phenomenon, which is very difficult for untrained human subjects to judge its existence. Therefore, more than just subjective experiments, carefully designed psychophysical experiments are required to collect sound experimental data and derive accurate color breakup prediction models.

In literature, the CBU-related studies can be categorized as follows. (a) Analytical method: The moving target is mathematically modeled by its colorimetric parameters and moving velocity [5]. Assuming perfectly pursuing eye movement, the perceived CBU is predicted by Grassman’s law of additive color, which unfortunately does not hold under eye movement, as shown in Figure 4.

7

Figure 4: (a) An ideal light emission patterns (not possible to realize). (b) Perceived original image. (c) Actual light emission patterns with positive and negative equalizing pulses.

(d) Resultant perceived image [5].

(b) Photonic measurement: High-speed cameras are used to emulate the eye movement and to capture the process of colors falling apart [6]. Such experiments build the basis of photometry-based analysis, which however is still discrepant from the visually perceived artifacts, as shown in Figure 5.

8

Figure 5: The target image ran at the 10° saccade. Initial and end views are marked with dashed rectangles [6].

(c) Subjective measurement: Commonly used in subjective experiments is a white box moving linearly on black background to provoke the worst CBU. The task of human subjects is to judge if any chromatic strips perceived on the edges [7]. In our experience such setup fails to reproduce reliable data because (i) perfectly pursuing the linear target movement on small displays is difficult and results in different degrees of CBU, and (ii) focusing on whether the edge is colored or not hinders us from considering the other important parameters such as stimulus size and spatial frequency as shown in Figure 6 and Figure 7.

Figure 6: The motion contrast measurement and analysis method [7].

9

Figure 7: Two different color transitions used to judge CBU [7].

(d) Psychophysical measurement: The foundation of Color Science is based on psychophysics because even when judging color of still stimuli the bios of human subjects can lead to significant variation. Thus, to accurately access the spontaneous CBU phenomenon, carefully designed psychophysical experiments are required, as shown in Figure 8 and Figure 9 [8].

Figure 8: Sequence of screen configurations for the saccade task. This is an illustration only – distances and sizes of objects on screen are not in correct proportions [8].

10

Figure 9: Illustration of the white bar, with or without a yellow and red color edge, used in the sequential color task. The bars are not drawn to proportion. The color edges are shown

in grayscale here, and are widened for easier viewing [8].

The goal of this chapter is to design an apparatus for assessing the detectability of color breakup for human subjects. Our psychophysical method is the Forced Choice method. In a forced choice experiment, two stimuli are presented and the human subjects have to pick the one with color breakup even if they cannot distinguish. In this way, the subjects’ preference of

“detected”/”not detected” can be filtered out.

We designed an experimental platform which can present stimuli with either sequential primaries or simultaneous primaries. In the sequential mode, which emulates a field sequential display, the red, green, and blue LED backlights are triggered one after the other with adjustable frequency, duty cycle, intensity, and order. In this mode, perceiving color breakup is possible. In contrast, in the simultaneous mode, which emulates a conventional display, the backlights are triggered at the same time, so color breakup is impossible to be observed. In both modes, the triggering frequencies are the same, so they both have the same degree of flickering. The subjects will not be able to use flickering as cue to guess, so we can separate the artifacts in the chromaticity domain (color breakup) from the luminance domain (flicker).

11

The stimulus is a vertical grating pattern moving along a circle (Figure 10). Adjustable parameters include color, contrast, speed (angular velocity and radius), size, and spatial frequency [9][10].

Figure 10: Grating is clear at 3/9 but blur at 12/6 o’clock.

Compared with linear movement, our method has the following advantages: (1) Circular motion is easier to trace. Otherwise, when the subject fails to trace the moving pattern, severe color breakup will be perceived. (2) The vertical grating generates different spatial frequency – DC at 3 and 9 o’clock, and maximum at 12 and 6 o’clock. Therefore, the subject is supposed to perceive different contrasts between 3/9 and 12/6 o’clock, and we can isolate spatial frequency from the other variables such as velocity or contrast.

The hardware is partially based on the previous work in [11][12]. The architecture is reviewed as follows.

CCFL BACKLIGHT

INVERTER Backlight Module

LED BACKLIGHT

LED DRIVER LCD PANEL

Figure 11: Block diagram of the proposed platform for measuring pursued color breakup.

12

A 19” TN-type LCD monitor (ViewSonic, VX912) was reworked for our purpose. Two lightbars, each with 24 LED chips, take place of the original CCFL backlights on the top and bottom edges of the panel. Three-in-one RGB LED chips (5WRGGB, Arima Optoelectronics Corporation) were used, as shown in Figure 12. To detour heat dissipation from the light-bar, the heat sink compound was applied to the gap between the thermal pads of LEDs and the lightbar, which attach to the metal frame tightly for better heat dissipation.

Figure 12: Close-up of LED chips with one red die, one blue die and two green dies and LED light-bar with 24 LED chips.

For efficiency and stability, current mirror was chosen to drive the LEDs because it can supply stable constant electrical current and adapt to different forward voltages of each LED.

It is suitable for driving LEDs because the output luminance is a function of the forward current through LEDs. A high constant current mirror (DD311, Silicon Touch Technology) was chosen as the LED driver because it can sustain up to one ampere forward current. It is a single-channel constant current LED driver incorporated current mirror and current switch.

The maximum sink current is 100 times the input current value set by an external resistor or bias voltage. The maximum output voltage of thirty-three volts can provide more power to LEDs in series. The output enable (EN) pin allows dimming control or switching power applications. Based on the above mentioned thermal and charactictics, the maximum forwarding current of LEDs can be determined so that the reference current can be derived by divided by one hundred. The LEDs’ voltage supply VLED can be determined by calculating each LED’s VTH in series connection (Figure 13-17).

13

Figure 13: Current mirror for driving LEDs.

Figure 14: Layout of LED driver and lightbar.

Figure 15: Schematics of PCB design.

Figure 16: The final PCB.

14

Figure 17: The spectra of RGB LED backlights. Red, green, and blue LEDs have peaks at 631nm, 535nm, and 460nm, respectively.

To process the video signals, we used the Altera Development and Education board (DE2, Terasic), which is based on the Cyclone II 2C35 FPGA. The on-board TV decoder (ADV7181B, Analog Devices) was used to decode the input composite video signals into YCrCb format. We modified the factory sample Verilog codes to manipulate images and perform desired image processing. The results were output by a digital-to-analog converter (ADV 7123, Analog Devices) to generate the 640x480 VGA signals. We also used the FPGA to control the backlighting patterns. The control signals were outputted from the GPIO interface to trigger the LED drivers. Pulse-width-modulation was used to adjust the backlight intensity.

To generate the stimuli, we used the Psychtoolbox, a Matlab-based library of handy functions for visual experiments [13]. It provides a simple interface to the high-performance, low-level OpenGL graphics library. To study color breakup under smooth-pursuing eye movement, our Matlab program is capable of generating moving Gabor patterns at different velocity, contrast, color, spatial frequency, and central/para-fovea area.

15