The LCDs have replaced the cathode ray tube (CRT) as commercial display products due to its advantages in weight, volume, and not radiation emitting. However, an LCD is not self-emissive, so it requires a backlight module as the light sources. Backlight module and LC layer are two main components while constructing an LCD. The structure of a TFT-LCD is illustrated in Fig. 1. A full-on white light source is lightened by the backlight module, uniformed by passing through a set of optical films and directed to transmit into the LC layer.
The TFT and color filters also reduce light transmittance since the total optical throughput is only about 7% of the total input. Especially, two third of light intensity form the backlight module would be absorbed by color filters.
Spatial color mixing
White B/L
Fig. 1. The structure of the TFT LCD.
The FSC-LCD [1], which is shown in Fig. 2, is a less power consuming display. Without color filters, it increases three times of light transmittance. The FSC-LCD sequentially flashes red, green, and blue field images with RGB LEDs to form a full color image, as shown in Fig.
3 [2]-[6]. Based on the temporal color mixing phenomena, human visual system will combine these images as a full color image. Without color filters, the FSC-LCDs are better than conventional LCDs with lots of advantages such as lower power consumption, higher transmittance, higher possible image resolution, and lower material cost. With these advantages, an FSC-LCD is no doubt a popular next-generation Eco-display.
Temporal color mixing
Blue Field Green Field
Blue B/L Green B/L
Red B/L Red Field
Fig. 2. The structure of the FSC-LCD.
Fig. 3. Temporal color mixing to form a full-color image.
Three main parts are necessary when it comes to driving scheme of FSC-LCDs: the TFT addressing, the LC response, and the backlight flashing. The timing chart is illustrated in Fig.
4; each frame is divided into three fields. The frame rate is 60 Hz as the conventional LCD,
flashing backlight. After all these steps, the inherent temporal color mixing mechanism in human visual system will generate a full color image.
Besides, public information displays (PIDs), as shown in Fig. 5, are now more and more popular that LEDs (a) and large size LCDs (b) are often applied to. The characteristics of LEDs are high luminance, high color saturation, but low spatial resolution. In contrast, large size LCDs have high resolution but low transmittance and color saturation. Since all PID applications require large size and high luminance, conventional LCDs reduce color saturation to earn higher luminance than usual; while the resolution can be enhanced by increasing the number of LEDs. However, reaching these demands would enlarge the cost. So FSC-LCDs here are more promising to be utilized to achieve higher resolution than LEDs and higher luminance than conventional LCDs.
Fig. 4. The timing chart in a frame time of FSC-LCD.
(a) (b)
Fig. 5. Public Information Displays (PIDs) applications on (a) scoreboards and (b) vending machines.
1.2 Color Breakup (CBU) Phenomenon
Without using color filters, field-sequential-color LCDs can perform three times higher optical throughput than conventional LCDs. Nevertheless, due to the way FSC-LCD displaying a colorful image is to sequentially display fields, a serious issue, color breakup (CBU), might come up with the relative speed between the displayed image and the viewer’s eyes [7]. CBU looks as a rainbow blur on the object fringe in the image as shown in Fig. 6.
The olor breakup phenomenon happen when there is relative motion, the separated colors on image edges can be perceived by human eye after visual integration. The CBU phenomenon would degrade image quality and cause discomfortable to human eyes. The mechanism and some prior resolutions of CBU will be further discussed in chapter 2.
(a)
Eyes Movement
(b)
Fig. 6. (a) Target image, and (b) simulated color breakup image with 10 pixels shifted.
1.3 Motivation and objective
Though FSC-LCD has three times higher optical throughput than conventional LCDs, it also has to drive at least three times faster in each pixel. Taking a 60Hz frame rate of RGB FSC method as an example, the field rate is 180Hz, which is only 5.56ms for each field. The length of TFT scanning is more than 3.3ms, and the backlight flashing length of each field is about 1.22ms. Therefore, the LC response time is limited to be shorter than 1ms. Several methods are brought out recent years to save as much time for LC response as possible, such as over-drive [8]-[9], and multi-division backlight [10]. The response time of the commercial LC modes are all longer than 5ms. Even the specific LC mode, nor can optically compensated bend (OCB) mode reach the response time. Thus, a two-color-field driving scheme [11], which allows longer time for LC response in FSC-LCDs and a two-field with two color filters
method (2F2CF) [12][13] were both proposed. However, the two-color-field method needs a large number of backlight divisions to maintain the image fidelity while the 2F2CF method lowers the light transmittance through the special color filters. The detail of these two driving scheme will be further elaborated in chapter 2. Hence, the objective of this thesis is to develop a 120Hz color-filter-less FSC LCD with high image quality, and less number of backlight divisions.
1.4 Organization
This thesis is organized as follows: In Chapter 2, the mechanism of causing CBU, some prior arts to solve CBU with multi-color-field, the two-field driving schemes, the color space transforming, the evaluation index for color difference and CBU suppression will be discussed. In Chapter 3, the proposed method, 120Hz Stencil-LPD method that can maintain the image fidelity without color filters will be detailed. The concept and algorithm of the proposed method will be illustrated. The simulation results with discussion will be presented at the same time. In Chapter 4, the optimization and the final results will be shown. Last but not least, the conclusion and future work will be given in Chapter 5.
Chapter 2
Principles
Color filter-less FSC-LCDs have the advantages of higher optical throughput, lower material cost, wider color gamut, and a possibility of three times higher resolution. However, color breakup phenomenon caused by the relative velocity between human eye and screen
Color breakup phenomenon correlates to the mechanism of human eye. Fig. 7 shows the cross-section of a human eyeball and a schematic enlargement picture of the retina [14]. The reflected light from objects is transmitted and refracted by the lens and eventually projected onto the retina. Retina is a nerve tissue in the eye that is responsible for sensation of light. It converts the optical signals of the visualized object to electrical impulses, which are then sent to the brain via the optical nerve. There are two types of photo-receptors in the human eye; the rods and cones. The rod cells are more abundant than cone cells in the retina, as shown in Fig.
8 [15]. Rod cells are sensitive to low levels of illumination (scotopic vision); the responses of which are slower and are not involved in color vision. On the contrary, cone cells are sensitive to high levels of illumination (photopic vision) with faster responses and produces color vision. Moreover, cone cells are further divided into of S-cones, M-cones, and L-cones according to their responding wavelengths, which are short-wavelength, middle-wavelength,
and long-wavelength respectively as shown in Fig. 9 [16].
Fig. 7. A drawing of a cross-section through the human eye with a schematic enlargement of the retina including rod cells and cone cells.
Fig. 8. A schematic drawing of rods and cones cell.
Fig. 9. Response of the three human cone types to the light of different wavelengths.
2.2 Color Breakup Mechanism
Color breakup (CBU) phenomenon occurs when human eye and the displayed object are moving under different speeds. This phenomenon results in not only eyes discomfort but also degrades image quality. Hence, understanding the mechanism of color breakup is necessary in order to come up with ways to inhibit the annoying phenomenon. As mentioned before, color breakup is dependent on the types of eyes movement, saccade or smooth pursuit. Therefore, considering this aspect, color breakup phenomenon can be classified into two types: dynamic color breakup and static color breakup, according to the moving feature of images and eyes movement.
2.2.1 Dynamic Color Breakup
Dynamic color breakup phenomenon mostly occurs on the fringes of moving objects on FSC-LCDs. Fig. 10 shows the perception mechanism of dynamic color breakup [17]. The horizontal axis represents the position in a FSC-LCD while the vertical axis shows time progressing. The colors comprising the white color, which are red, green, and blue, are displayed temporally in one field time. Through eye integration, the three primary-colors are projected and separated onto the retina and results in a rainbow-like color bar on the edges of
the white image, which is then perceived by human eyes. Through eye integration, the three primary-colors are projected onto the retina and a rainbow-like color bar occurring in the edge of white image is perceived by human eye.
Fig. 10. The mechanism of perceiving dynamic color breakup.
2.2.2 Static Color Breakup
Static color breakup occurs when human eyes glance at a still image. Fig. 11 shows the mechanism of perceiving static color breakup [17]. The fixed white image which consists of the three primary-colors, R, G, and B, is displayed in a FSC-LCD. When human eyes glance at the image to obtain detail features, the R, G, and B sub-field image are projected onto the
Fig. 11. The perception mechanism of static color breakup.
2.3 Color space transformation and evaluation index of color difference 2.3.1 Color space transformation
Considering the concept of trichromatic color space, every color is mixed by the three primary colors: red, green and blue. Fig. 12 shows the setup of the color-matching experiments [17]. An arbitrary light of the test colors illuminates the lower half of the white screen and produces a visual stimulus to the human eye through the black shadow. Meanwhile, the upper half of the white screen is illuminated by a light source that consists of the three primary colors. The intensities of the red, green and blue light sources are adjusted respectively in order to match the color projected on the lower screen. We see that it is possible to find a set of R, G and B that matches any specific color. In 1931, the CIE used the
three primary colors with wavelengths of 700nm, 546.1nm and 435.8nm to match all visible monochromatic lights as shown in Fig. 13 (a), and the CIE 1931(R, G, B) chromaticity diagram can be obtained as shown in Fig. 13 (b).
Fig. 12. Experimental setup of color-matching experiments.
(a) (b)
Fig. 13. (a) Tristimulus values for different wavelengths, and (b) CIE 1931 (R, G, B) chromaticity diagram.
Since different color stimuli can be obtained by linear summation, the spectrum of
colors cannot be acquired no matter how the (R, G, B) intensities are adjusted. To resolve this issue, we can move one of the primary sources from the upper screen to the lower screen on top of the desired color. As Fig. 13 (b) shows, the negative r values are sometimes not suitable for describing particular colors [19]. To solve this problem, the CIE1931 (X, Y, Z) system was proposed with
Eq. 1, as shown in Fig. 14. CIE 1931 (X, Y, Z) chromaticity diagram.By using this method, all colors can now be described with positive values. One important feature of the CIE1931 (X, Y, Z) system is that the Y value was set as luminance of the stimulus in terms of lm sr-1 or cdm-2. To get the CIE 1931 (X, Y, Z) values from a spectrum, each wavelength is summed linearly as shown in
Eq. 2.
Fig. 14. CIE 1931 (X, Y, Z) chromaticity diagram.
Eq. 1
Eq. 2
where k=683 lm W-1, which represents the transformation from radiometry units (W) to photometry units (lm), and P(λ) is the spectral distribution of the stimulus in terms of W sr-1 m-2.
Although the CIE 1931 (X, Y, Z) color system can precisely define a color, there is a problem when dealing with color difference and its tolerance [20][21]. Fig. 15 shows the well-known MacAdam ellipses in the CIE 1931 (X, Y, Z) chromaticity diagram [22]. Human eye would not distinguish color difference within the ellipses in this figure. To better illustrate
Eq.
3 derives the coordinate transform from the CIE 1931 (X, Y, Z) to the 1976 (L*, a*, b*) color system. In the equation, Xn, Yn, and Zn are the three tri-stimulus values of reference white. L* represents lightness, a* approximate redness-greenness, and b* approximate yellowness-blueness. Therefore, the color difference is evenly given by the formula of ΔE*ab. The CIELAB provided a uniform chromaticity diagram, as shown in Fig. 16, which makes most of color difference formulas were established based on the CIELAB color space.
Fig. 15. MacAdam ellipses in the CIE 1931 (X, Y, Z) chromaticity diagram.
Fig. 16. CIELAB color space.
Where
Eq. 3
The CIE1976 (L, U, V) color space was adopted as the CIE 1931 (X, Y, Z) color space with perceptual uniformity. It is extensively used for applications such as computer graphics
system. This color space is employed to determine backlight signals in this thesis due to its uniform characteristic (Fig. 17).
Fig. 17. CIE1976 u’v’ chromaticity diagram.
Eq. 4
2.3.2 Color difference of CIEDE2000
Based on the color difference of 1976 CIELAB, CIE proposed the CIEDE2000 to modified the formula to describe all color difference ranges [23]-[26]. To revise the issue of
color uniformity, the formula, as shown in
Eq. 5, considers the weighting function of lightness (SL), chroma (SC), and hue (SH), which are
shown in
Eq. 6. The kL, kC, and kH values are the parametric factors, which are adjusted under different viewing parameters, for the lightness, chroma, and hue components, respectively. RT
function intends to improve the performance of the color difference equation for describing
chromatic differences in bluish colors.
Eq. 5
Eq. 6 Where
2.4 Prior Color Breakup Suppression Methods
Color breakup phenomenon is a fatal drawback of FSC-LCDs. This phenomenon causes discomfort in human eyes and degrades image quality. Therefore, how to resolve the color breakup issue has been a major research in the FSC technique. Many color breakup suppression methods were proposed in different ways which can be categorized into motion compensation, inserting fields, and reducing color difference between sub-frames.
2.4.1 Motion compensation
The first part, motion compensation, was proposed to solve the dynamic color breakup.
suppressed. However, if some minor objects having a chance to cross the observer’s tracking path in the opposite direction, the perceived image of the crossing object would get worse than the conventional RGB driving method, as shown in Fig. 19.
Fig. 18. The mechanism of motion compensation; color breakup was suppressed effectively.
Fig. 19. The mechanism of motion compensation; object and observer’s eye tracing path are in the opposite position.
2.4.2 Mono-color fields
In the second part, inserting fields, it can be further broken into two sections, mono-color fields and multi-color fields. Inserting mono-color fields is to increase the field rate up to 360Hz or higher, such as RGBRGB or RGBKKK [27][28]. By doing so, while there is eye movement, the width of separated color bars on the border of the displayed image would be reduced, as shown in Fig. 20. This method is often applied in the DLP projector with color wheel and digital micro-mirror-device (DMD). Moreover, inserting complementary color
(a) 180Hz RGB (b) 360Hz RGBRGB
Fig. 20. Simulation color breakup image of (a) 180Hz RGB, and (b) 360Hz RGBRGB.
(a) 180Hz RGB (b) 300Hz RGBCY
Fig. 21. Simulation color breakup image of (a) 180Hz RGB, and (b) 300Hz RGBCY.
2.4.3 Multi-color fields
The solutions mentioned above were proposed to effectively suppress color breakup, yet they are limited to be implemented on hardware by the strict LC response. In order to accomplish FSC method on hardware, the field rate must be lowered. Consequently, the Stencil-FSC methods [29]-[31] and the LPD method [32][33] with multi-color fields were proposed to make field rate lower than 240Hz and suppress color breakup effectively.
(a) Stencil-FSC methods
In Fig. 22, it shows the driving schemes of Stencil-FSC methods under different field rate. The main idea of Stencil-FSC method is to provide multi-color fields using locally controllable and dimmable backlight module. The 240Hz Stencil-FSC method is to display major information of input image with high luminance and rough color in the first field. Then
the other three fields show the detail of the remaining information in the three primary colors with lower luminance. Thus, luminance of the perceived separated color bars on the edge of the image would be concentrated in the first field, which means the color breakup phenomenon is effectively suppressed and almost imperceptible. In Fig. 23, we compare the color breakup phenomenon between conventional RGB driving method (Fig. 23(a)) and the 240Hz Stencil-FSC method (Fig. 23(b)).
The field rate of the 240Hz Stencil-FSC method is still a little higher than RGB driving method, so our team further proposed the 180Hz Stencil-FSC method to solve this issue. The concept of the 180Hz Stencil-FSC method is to combine the all green information with some parts of red and blue information in the first field. Then the rest of red and blue information are displayed in the second and the third field. This is due to human visual sensitivity acts stronger in the green color than in the other two primaries. In this regard, when color breakup comes up, the separated colors on the fringe do not contain green information so that the rainbow-like fringe can hardly be observed, as shown in Fig. 23(c).
(a) Color breakup image of RGB-driving method with target image Girl.
(b) Color breakup image of 240Hz Stencil-FSC method with target image Girl.
(c) Color breakup image of the 180Hz Stencil-FSC method with target image Girl.
Fig. 23. Color breakup of image Girl by different FSC methods: (a) RGB-driving, (b) the 240Hz Stencil-FSC method.
(b) LPD method
The concept of LPD method is to redistribute the original three highly saturated primary-color backlight fields into less but sufficient ones, as shown in Fig. 24. In Fig. Fig.
24(b), the right hand side shows the three desaturated-primary-color fields, and Fig. 24(c) shows the reproduced image with color breakup phenomenon. Reducing color saturation can
lower the color difference between each two fields that the separated color would just be obtained as motion blur but not color breakup. To precisely reproduce an image as the target, the LPD method also utilizes the locally controllable backlight module as the Stencil-FSC methods. In the LPD method, the pixel chromaticity distribution of each backlight segment is analyzed in the CIE1976 u’v’ uniform color space, as Fig. 25 shows. By doing so, the colors of the three backlight fields are desaturated by mixing other colors in the same field that the color breakup was significantly reduced.
Fig. 25. The triangles represent color saturation of sRGB and LPD backlight primary-colors in the CIE1976 u’v’ uniform color space.
(c) Two-field methods
The 180Hz methods can be implemented on the OCB mode of LC displays but still not enough for large size or commercial LC modes. Accordingly, the two-field method with two color filters (2F2CF) and the 120Hz two-color-field method were proposed to be applied to any mode of commercial LC and any size of LCDs.
A spatial-temporal display method, which has 2 fields and 2 color filters that is called 2F2CF method, was proposed to enhance image quality of displays [12][13]. This method combined spatial color mixing with color filters and temporal color mixing with sequential backlight flashing. Each pixel has two sub-pixels with two different color filters and two fields with different backlight colors. The idea of 2F2CF method with two typical types is illustrated in Fig. 26 and Fig. 27 [12]. By changing the spectra of the color filters and backlights, different color combinations can be achieved.
Fig. 26. A 2F2CF type with yellow-cyan (Y-C) backlights and green-magenta (G-M) color filters. (a)Spectrum-level illustration for the color synthesis procedure. (b)The configuration of type-I 2F2CF LCD.
Fig. 27. Another 2F2CF type with green-magenta (G-M) backlights and yellow-cyan (Y-C) color filters. (a)Spectrum-level illustration for the color synthesis procedure. (b)The configuration of type-II 2F2CF LCD.
Fig. 27. Another 2F2CF type with green-magenta (G-M) backlights and yellow-cyan (Y-C) color filters. (a)Spectrum-level illustration for the color synthesis procedure. (b)The configuration of type-II 2F2CF LCD.