The image frame time is separated into at least three intervals for red, green and blue as shown in Fig. 1-5. Each interval is for the rotation of LC and backlight emission. FSC-LCDs can then display a field image correctly at each field. Therefore, FSC-LCDs need a fast liquid crystal response time to rotate to the target value in a field time, which is about 1/3 frame time if there are three fields in one frame. If the response time of liquid crystal is not fast enough, the blocked volume of light will be not correct [7]. After human eye integration, the hue, saturation, intensity of color might be different when compared to the input image. This phenomenon is called color shift. Although LEDs have a very wide color gamut, a slower liquid crystal will shrink the color gamut and display the incorrect color.
Time
Backlight duty time
TFT scanning
a
b c
One frame time
Fig. 1-5 Three intervals for red, green and blue fields
a: TFT addressing time; b: time of LC rotating; c: backlight flashing time
1.4 Motivation and Objective
FSC-LCDs have higher optical throughput without color filter requirement which saves more power and creates a wider color gamut using multi-colored high power LEDs. However, FSC-LCDs still suffer from the color breakup artifact when
4
5
human’s eyes follow saccade or smooth pursuit movement across the screen. Our group had proposed a Dominant-Field plus Three Primary Fields called DRGB to effectively suppress the CBU artifact [8]. However, the concept of DRGB method needs a precise and complicated calculation process. FSC-LCDs use DRGB method without frame buffers is hard to implement in hardware. Therefore, the objective of this thesis is to propose a modified DRGB method (called Fast-DRGB [9]) to simplify the calculation process and implement the Fast-DRGB method in hardware without any frame buffers.
1.5 Organization of This Thesis
This thesis is organized as follows. In Chapter 2, some prior arts to suppress CBU and characterize color shift phenomenon in FSC-LCDs are presented. In Chapter 3, the concept of the Fast-DRGB method is proposed to implement on field
programmable gate arrays (FPGA) with a 15.4” panel. In Chapter 4, simulation of the Fast-DRGB will be discussed. In Chapter 5, the hardware implementation of the Fast-DRGB and panel characterization with modified color model will be demonstrated. Finally, conclusions and future works will be given in Chapter 6.
6
Chapter 2
Prior Arts in Field-Sequential-Color LCDs
FSC-LCDs have advantages of higher optical throughput, lower material cost without color filters requirement, a wider color gamut, higher resolution without subpixels. However, the mechanism of eye movement on FSC-LCDs degrades the image quality because of CBU artifact. The LC response time is not fast enough to cause color shift phenomenon. Many researches on CBU artifact suppression and color shift characterization in FSC-LCDs have been discussed in recently years.
2.1 Mechanism of Color Breakup (CBU)
2.1.1 Human Eye Structure and Eye Movement
In the beginning, we should discuss the human color vision. The image use light to be transmitted and refracted by the lens, finally projected onto the retina. The structure of eye ball is depicted in Fig. 2-1 [10]. The retina is composed of photosensitive cells, rods and cones and is shown in Fig. 2-2 [11]. These cells transducer the image information to optic nerve and then brain receives the signals to decode signals into a color full image. The retina contains two kinds of photoreceptor for receiving the image details.
Fig. 2-1 The cross section of eye ball
The rods are in charge of achromatic color at low luminance (scotopic vision) while the cones are in charge of chromatic color at high luminance (photopic vision).
And both cells work in the intermediate luminance environment (mesopic vision).
Fig. 2-2 A schematic drawing of rods and cones cell
After discussing the human vision system, two kinds of eye movement are related to color breakup issue, saccade and smooth pursuit [13][14][15]. Saccade is human eyes jump from one position to another position and the velocity, direction of eyes is voluntary.
7
(a) (b)
Fig. 2-3 (a) Perceived image, (b) trajectory of saccadic movement
Fig. 2-3(b) shows that lines are the trajectory of saccadic eye movement. After several cycles of movement, the clear image will be perceived and is shown in Fig. 2-3(a) [12]. Smooth pursuit is human eyes follow the moving object at the same velocity, so that the image is clearly when eyes are doing smooth pursuit.
2.1.2 Two Types of Color Breakup
There are at least three fields (Red, Green and Blue) needed to make a correct full color image in FSC-LCDs. Because of the temporal color mixing, the frequency of alternation between three fields should be quicker enough so that different field images are projected in the same position of retina. The brain will receive a correct image. If the projected images are not in the same position of retina, human eyes perceive color breakup (CBU) phenomenon. CBU is strongly dependent on two types of eye movement: saccade and smooth pursuit movement. Therefore, two kinds of CBU phenomenon are discussed, static and dynamic.
8
2.1.2(a) Static Color Breakup
When a static image is perceived during saccadic eye movement, the CBU happens along the eye movement direction and is illustrated in Fig. 2-4. The left dash block means image is still and only eyes are moving. Other three blocks show three primary field images locate on different positions of retina cause static CBU phenomenon.
Red field Green field Blue field
Fig. 2-4Saccadic eye movement cause static CBU
2.1.2(b) Dynamic Color Breakup
When tracking a dynamic image, the human eye will pursue the moving image as shown in Fig. 2-5. When white bar is moving to specific direction, human eyes will chase white bar through smooth eye pursuit. Human eyes will detect different color fields combination on the edges of white bar after the color temporal mixing integration.
9
Display area
Moving velocity : V
Perceived image
Fig. 2-5 Smooth pursuit movement on moving object causes CBU
2.2 Evaluation of CBU –
Color Difference Equation CIEDE2000 (ΔE
00)
When describing how human eyes perceived the color, the researches of International Commission on Illumination (CIE) are first published mathematically.
CIE defined color space from a series of psychological experiments. In 1920s, W.
David Wright and John Guild designed experiments to measure color matching functions. In 1931, the Colorimetry Committee of CIE selected three primary wavelengths (435.8nm, 546.1nm and 700nm) and 17 standard observers to experiment in color matching functions. These experiments are under specific conditions like a bipartite area subtending a 2 degree viewing angle with dark environment.
10
However, the measured color matching functions (r(λ),g(λ)and b(λ)) are complicated owning to negative values shown in Fig. 2-6(a). After a mathematical transformation, the CIE defined color matching functions (x(λ),y(λ)and z(λ)) are all positive values for simply calculation shown in Fig. 2-6(b) [16]. Using transformed color matching functions, CIE defined a color space with tristimulus values X, Y and Z called CIE XYZ. The CIE XYZ color space has conditional effect, so the light source (L) and reflectance (R) factor should be considered to derive X, Y and Z values as shown in Eq. 2-1-1.
∫
= L λ R λ x λ dλ
X ( ) ( ) ( ) (Eq. 2-1-1)
∫
= L λ R λ y λ dλ
Y ( ) ( ) ( ) (Eq. 2-1-2)
∫
= L λ R λ z λ dλ
Z ( ) ( ) ( ) (Eq. 2-1-3)
(a) (b)
Fig. 2-6 (a) Original color matching functions, (b) Modified color matching functions
11
Furthermore, the curvature of y(λ)is very similar with human luminosity function. Therefore, CIE matches y(λ) to human luminosity function with multiplying a coefficient, and Y parameter describes brightness or luminance of color.
The chromaticity of color can be described as x and y, the normalized values from CIE XYZ color space in Eq. 2.2-1. The derived color space specified by x, y and Y and is known as CIE xyY which is shown as Fig. 2-7(a).
However, CIE XYZ or CIE xyY color spaces are not uniform to describe color difference correctly. For examining color quality, different color spaces such as CIELAB, CIELUV had been proposed to quantify the color difference and is shown in Fig. 2-7(b) [16] and Fig. 2-7(c). The CIELAB color space is the uniform color space and most close to human opponent vision system. The transformation from CIEXYZ to CIELAB is non-linear are described in Eq. 2.2-4. The LAB means lightness (L*), color component of red-green (a*) and color component of yellow-blue (b*) respectively.
12
16
(c)
Fig. 2-7 (a) CIE xyY, (b) CIELAB, (c) CIELUV color spaces
Furthermore, in 2001 the CIEDE2000 color difference formula [17] based on CIELAB color space formula is published. The formula compared the standard and the sample using the lightness (L*), chromaticity (C*) and hue (H*) which are derived by Eq. 2.2-8. Then ΔL*, ΔC* and ΔH* calculate the color difference. ΔR term is a error compensation between chromaticity and hue, and the SL, SC and SH are weighting coefficients for lightness, chromaticity and hue respectively. The KL, KC
and KH are the factors of different viewing conditional parameters, like textures, backgrounds, etc.
15
The main purpose of CIEDE2000 formula is to compare color difference between original image and simulated image. CBU phenomenon happens when human eyes observe the rainbow edge. If the color difference area between rainbow edge of simulated CBU image and same position of original image is large, the CBU phenomenon is apparent. Otherwise, if color difference is small the CBU phenomenon is less observed. Therefore, using CIEDE2000 formula to calculate the color difference between rainbow edge of CBU image and the same position of original image and then the CBU phenomenon level is estimated. With CIEDE2000 value, we can compare different simulated CBU images using different methods to estimate how effectively these methods suppress CBU phenomenon.
2.3 Prior Arts to Suppress CBU Phenomenon
The researches on suppressing CBU phenomenon have been improved recently.
Different kinds of CBU suppression methods are shown in Fig. 2-8, such as Mono-Color Field, Multi-Color field, and Motion Compensation.
In Mono-Color Field part, the field rate is increased to 360Hz or higher (RGBRGB or RGBKKK) [18][19], inserting complementary color fields to original three primary fields (RGBCY). Above methods are trying to decrease distance of field images on retina while eye movement as simulated in Fig. 2-9.
Mono‐Color Field
Field Rate Increasing
Primary Color Field Insertion
Multi‐Color Field
Primary Color Mixing
120Hz (2‐field)
RB+GB (2‐field) RGBRGB (6‐field)
RGBW (4‐field) RGBCY (5‐field)
RGBKKK (6‐field)
FSC LCD
Stencil‐FSC
240Hz (4‐field) 180Hz (3‐field) RGBD (4‐field)
Motion Compensation
4‐Color Field
Arrangement(4‐CFA) 240Hz (4‐field) Adjust of Color
Element (ACE) 180Hz (3‐field)
Fig. 2-8 Different kinds of CBU suppression methods in FSC LCD
16
17
RGBKKK 360Hz RGBRGB
360Hz
RGBCY 300Hz
Fig. 2-9 Increasing field rate or adding complementary color to suppress CBU
About Motion Compensation, the dynamic color breakup is suppressed effectively. N. Koma was proposed “Adjust of Color Element on the Eyes” (ACE) [27] and is shown in Fig. 2-10. The ACE method displays red, green and blue fields on different positions of retina when eyes are moving. Therefore, human eyes observe the CBU less image with ACE method. However, if human eyes are moving to positive direction and the ACE method displays negative direction, the CBU phenomenon is observed largely, as shown in Fig. 2-11.
Fig. 2-10 (a) Conventional and (b) motion interpolation method in FSC-LCDs
Fig. 2-11 The ACE method displays red, green and blue images from left to right and observer’s eyes pursuit the image from right to left.
The Four Color Field Arrangement (4-CFA) of Motion Compensation was proposed by Ya-Ting Hus [28] for suppressing dynamic CBU. The color sequence is displayed as RGBR, GBRG and BRGB in three continuous frames orderly at 240Hz and is shown in Fig. 2-12 (a). This combination of red, green and blue images make the moving edge to be white and the dynamic edge is eliminate, as shown in Fig. 2-12 (b).
18
(a)
(b)
Fig. 2-12 (a) The color displayed order and (b) the temporal color mixing with 4-CFA method.
Our group further improves the Primary Color Field Insertion of Mono-Color Field. Using 360Hz or higher frame rate is hard to achievable for hardware, because of being constricted to the current technology of liquid crystal response time. The 240Hz frame rate or less is achievable after upgraded the commercial twist nematic (TN) mode LC. Thus, at 240Hz frame rate the WminRGB (minimum white extract from three primary colors) method [20], DRGB (dominant color field of input image plus three primary) method [9] using global backlight.
Our group also proposed the Multi-Color Field, the Stencil Field-Sequential-Color Method [21] using local backlight. These methods greatly suppressed CBU phenomenon. Furthermore, 180Hz stencil method [22] and two-field-sequential method at 120Hz [23] were also proposed for different
19
applications on different panel size.
The main concept of these methods is putting the majority of image contents in the first field, and the intensities of three primary fields are lowered down and are shown in Fig. 2-13. Then four field images combine together to make a full color image. Human will be less sensitive to CBU edge during saccadic eye movement on still image or smooth pursuit movement on moving object and is illustrated in
First field Second field Third field Four field
RGB
@180Hz
none
WminRGB
@240Hz
Stencil
@240Hz
DRGB
@240Hz
Fig. 2-13 First field contains majority image so than decreasing intensities of rest fields
At 240Hz driving methods are mentioned above, WminRGB and DRGB methods are using global backlight; Stencil method is using local backlight. Because the constriction of hardware specification in 15.4” panel, the support backlight is global, the Stencil method is not included in our choices. Comparing WminRGB and DRGB
20
methods, WminRGB method always displays white backlight in the first field and DRGB method can display majority color in the first field depending on input image.
Therefore, WminRGB method averagely consumes more power when comparing with DRGB method does.
RGB @180Hz
WminRGB @240Hz DRGB @240Hz
Fig. 2-14 Simulated CBU images from 3 different methods
2.3.1 Concept of the DRGB Method
Ideally, total optical throughput can be simplified into a relation of backlight and digital gray scale value, which is described in Eq. 2-3-1,
21
throughput optical
total i
T
BL* ()= (Eq. 2.3-1)
Where T(i) is the transfer function from digital gray scale values i to transmittance of liquid crystal and T -1 is the inverse function. The gamma curve is shown in Fig. 2-155.
Between digital values and transmittance should be considered to maintain the white balance.
i
Transmittance
Digital Gray level T(i)
Fig. 2-15 Gamma curve function is related between digital value and transmittance
The DRGB method finds the majority of image information in dominant field according to input image. So, the dominant field image is different image by image.
The algorithm of DRGB method is using adaptive backlight to determine the dominant field backlight color (D-field backlight color). According to total optical throughput of the image in conventional LCDs, we can deduce the digital gray scale values and backlight gray scale values in FSC-LCDs by the equation. In the D-field, the gray scale values of backlights are set as BLr, BLg and BLb and backlight on rest of three primary fields are full on. After the D-field backlight (BLr, BLg, BLb) is determined by the DRGB algorithm, Eq. 2-3-2 shows the additional digital gray scale values d is calculated and then r’,g’,b’ are derived in the following.
22
))
The feedback control is used on determining the D-field backlight color from backlight choices. When image is loaded into the algorithm, the image is transformed to CIELUV color space and down sampled in 2-by-4 sampling rate for decreasing calculation. Then criterion E is calculated between original input image and simulated CBU image and because there are eight sets of (BL , BL BL ) in first feedback control, there are eight E values coming in first calculation. Finding the smallest one from eight E values and then keeping shrink the range of backlight choices. After 3 bit accuracy feedback control approach, the final backlight sets are applicable D-field backlight color for the input image to suppress CBU artifact. The flowchart of DRGB is shown in Fig. 2-16.
Δ 00
r g, b
Δ 00
Δ 00
23
input image
i sorted from 1(min) to 8 (max) A1
i sorted from 1(min) to 8 (max) A1
Fig. 2-16 (a) Flowchart of DRGB algorithm, (b) 3-bit accuracy approach
Because DRGB method needed the whole image contents three times for 3 bit accuracy feedback control approach. The frame buffers are need to storage the whole image contents for calculation. If we want to implement on 15.4” panel without frame buffers, the method can be utilized. Besides, the transformed color space CIELUV is complicated. So the proposed method Fast-DRGB method will simplify the calculation of the DRGB method for hardware implementation and with almost identical CBU suppression effect.
The panel supported by CPT uses commercial upgraded TN mode LC, so the response time of LC is not fast for ideal FSC-LCDs, and then color distortion might happen. In the following section, the factors causing color distortion are discussed.
2.4 Mechanism of Color Distortion
FSC-LCDs are well known to have a wider color gamut by using multi-colored LEDs as a backlight source instead of using color filters to mix color. Therefore, the saturations of three primary LEDs are very high so that increase the 3 apexes of triangle in the color gamut and is shown Fig. 2-17. [24]
Fig. 2-17Color gamuts with different backlight sources in CIE xyY color space
However, as mentioned before FSC-LCDs need at least 180Hz frame rate to make a color image. Thus, the LC response time is needed higher than 180Hz so that LC can be rotated to the specific position in each field. When LEDs emit through LC, the color of image is correct. Optically Compensated Bend (OCB) mode LC is most used in FSC-LCDs for its fast response time. However even the OCB has fast response time LC, they still have color distortion issue because of the structure of thin film transistor (TFT) arrays.
25
26
Furthermore, OCB mode LC is not popular in commercial product. Twist nematic (TN) mode LC is still mainly applied on LCDs by companies. After upgrading TN mode, different kinds of functional TN mode LCs are created such as speed-up TN mode LC. Although the response time of speed-up TN is not fast than one of OCB mode, the TN mode LC has enough time to rotate to certain position with modified driving methods on circuit board.
However, TN mode LC is not fast enough for the proposed Fast-DRGB method to implement on hardware. So when rotating the liquid crystal in dominant field, the target position is hardly achieved. The chance of achieving target position depends on how much time gray to gray requires. And the situation is the same in the rest three primary fields, the finally presented image is not correct.
2.5 Prior Arts to Characterize Color Distortion
The color distortion on OCB mode seems irrational, the reason is the capacitance multiple voltage is constant. Figure 2.18 (a) shows that setting the voltage from start to target value, and later the voltage will follow the CV=constant curve to make the target value wrong. Therefore, a method called overdrive was proposed to solve the issue as shown in Fig. 2-18 (b) [25]. Setting the target value higher than estimated value, the correct voltage will be achieved after stabilization in LC.
(a)
(b)
Fig. 2-18(a) Target voltage is unachievable in one frame time, (b) Target voltage is achievable using overdrive method
27
Overdrive method needs to give higher voltage in the beginning. If the voltage is too large and LED backlights cannot afford it, the LED backlights are easily burned-out.
The other method is building a color model to characterize color distortion issue.
The model consists of two stages. First is a linear transform from radiometric scalars to tristimulus values. Equation 2.5-1 shows the basic form of linear transformation.
Equation 2.5-2 shows more precious and complicated transformation. The symbol N means the accuracy bit number of display system and dR is the digital signal value of red field. The second stage is non-leaner transform between digital signal values in each color field and each radiometric scalar. These models are GOG, S-curve, and polynomial and are shown in Eq. 2.5-3, Eq. 2.5-4 and Eq. 2.5-5 [26]. Symbol i means the each radiometric scalar (R, G, and B). These models derive the new digital signal values. Then, inputting the new digital signal values into panel can measure expected tristimulus values.
GOG model:
Device independent and each field channel independent are prior calibrations to build the color model. Because the new digital signal values are derived from linear
Device independent and each field channel independent are prior calibrations to build the color model. Because the new digital signal values are derived from linear