The factors affect FSC-LCD’s performance are two. One is a mismatch of the LED resolution and the LC resolution. The other is light leakage from the neighbor divisions. Based on the two reasons, this thesis will incorporate the hue for the classifier. In the following sections, this thesis will introduce the hue, the details of the construction of the classifier, and the pictures chosen for simulation.
4.1.1 Hue
The hue which describes the kind of color such as red, green, and blue was from the Munsell color system [23]. The system was developed originally by an American artist, Albert Munsell in 1905, and the scale was refined and renotated by the Colorimetry Committee of the Optical Society of America (OSA) in the late 1930s.
The Munsell color system uses cylindrical coordinates to specify colors as shown in
31
Fig. 4-1 (a). The coordinates are hue as the circumferential angle, value (lightness) as the ordinate, and chroma as the radius.
The Munsell color system selected red (R), yellow (Y), green (G), blue (B), and purple (P) as five principle hues and spaced the principle hues equally around the hue circle. The colors YR, GY, BG, PB, and RP are inserted between each two principle hues. Moreover, the Munsell color system further divided the space between a principle hue and an intermediate hue into ten hues. Totally, a hundred hues were established. Forty hues are shown in Fig. 4-1 (b). The numerical calculation of hue is described by Eq. 4-1 and Eq. 4-2.
(4-1)
(4-2)
(a) (b)
Fig. 4-1 The Munsell color system: (a) the color solid for the Munsell color system, and (b) the Munsell hues [24].
4.1.2 The Construction of the Picture Classifier
To build a picture classifier suitable for FSC-LCD, knowing the issues which affect the image quality is necessary. In FSC-LCD, because the LED resolution is
32
always much lower than the LC resolution, a LED backlight signal needs to take care of many LC signals. The sequence of deciding the BL signal and the LC compensated signals is shown in Fig. 4-2. A green image is taken as an example. The location of LED backlights divides the image into divisions. In Fig. 4-2 (b), the LED backlight in each division is decided by the average, the square root of average, or the one-third power of the pixels in the division. By using Eq. 4-3 the LC compensated signals are calculated.
Thus, one of the issues affecting the image quality occurs when the colors in a division diverge greatly. If the colors in a division diverge greatly, the LED backlight will not be suitable for every pixel in either LED decision way and result in bad image quality. Another issue may occur in the blurring part. When the LED backlight in a division differs from the LED backlight in a neighbor division a lot, the compensated LC signal may not well compensate the light leakage from the neighbor divisions and result in bad image quality. Both the issues are related to color differences with the neighbors.
Based on the color differences, the thesis simplifies the images into simpler diagrams owning color difference information as Fig. 4-3 shows. In Fig. 4-3 (a), the RGB coordinate is transformed to the hue circumferential angle which represents the kind of color. Each pixel in the image has its neighbors. The red center represents the pixel and the yellow neighbors represent the red center’s neighbors as Fig. 4-3 (a) shows. Next, we average the differences between the yellow neighbors and the red center for the color difference characteristic of the red center. Thus, each pixel has its color difference characteristic. Finally, a histogram of the pixels is the simpler diagram owning color difference information.
Ten examples of the images and the corresponded diagrams are shown in Fig. 4-4.
33
In Fig. 4-4, according to different color difference information, each image has its own diagram. This thesis utilizes the average and the standard deviation as two standards to sort the data base. The sorting result is shown in Fig. 4-5. The bottom two figures are with low standard deviations while the top two figures are with high standard deviations. The left two figures are with low averages while the right two figures are with high averages.
(4-3) where
The I represents the image intensity, the BL is the backlight intensity, and the LC is the LC signal.
(a) (b) (c) (d) Fig. 4-2 The sequence of deciding the BL signal and the LC compensated signals: (a)
the target image and the divisions, (b) the BL signals, (c) the blurred BL, and (d) the LC compensated signals.
(a) (b)
Fig. 4-3 The construction of the picture classifier: (a) the target image and the divisions, and (b) the simpler diagram owning color difference information.
RGB Hue Yellow neighbors Histogram
Red center
34
Fig. 4-4 The examples of the images and the corresponded diagrams.
Fig. 4-5 The sorting result from the picture classifier: the horizontal axis is the average, and the vertical axis is the standard deviation.
4.1.3 The Pictures Chosen for Simulation
Besides the picture classifier, this thesis also consults the lightness which is an 0
0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01
0 0.05 0.1 0.15
hue
Average
Standard deviation
35
issue affecting the image quality in a polarizer-free display, the human face, the skin tone, and the sunset. Eventually, thirteen pictures were chosen including colorful images, plane images, sunset images, and images with skin as Fig. 4-6 shows. In the top column, from left to right are Canoe, Harbor, Soccer, and Candle. In the middle column, from left to right are Palace, Desert, Airplane, and Motorcycles. In the bottom column, from left to right are Girl, Cyan_moon, Beach, Pharos, and Woman.
Fig. 4-6 Thirteen pictures were chosen from the picture classifier.