Laboratory Practice I

Part II: LCD Color Characterization

Yousuf Hemani, 277149

Tanmay Mondal, 277157

Mohammad Al Lakki, 277153

March 24, 2016

Department of Physics and Mathematics

University of Eastern Finland

Tanmay Mondal Project Report, 29 pages

Yousuf Hemani University of Eastern Finland

Mohammad Al Lakki MS Photonics

Supervisors: Dr. Hannu Laamanen

Abstract

It is very important to achieve consistent and high quality of color reproduction in an

imaging system. This requires a clear understanding of the color characteristics of various

devices in the system. This can be done using device characterization, which means to

find out the relationship between the device coordinates and color coordinates. For every

display a certain kind of model is to be developed to check its performance. This whole

process defines a color management system for various digital displays on which users

heavily rely on for creating, viewing and presenting different color images. LCD displays

are becoming increasingly common due to certain specific advantages which are compact

size and low power consumption; increasingly being needed in laptop computers and

projection systems. In this report we emphasize on different models of color

characterization of an LCD. We compare the linear model and the masking model and

their results. The masking model is known to be better in dealing with the specific

limitations of the linear model when dealing with LCD systems such as channel

interaction and non-constancy of channel chromaticity. After taking characterization

measurements and optimizing the code, we convert the final values to CIELAB system

and calculate color difference for both the models which clearly shows that masking

model is better than linear model.

Table of Contents 1 Introduction ............................................................................................................................................... 4

2 Theory ........................................................................................................................................................ 6

2.1 Colour coordinates and chromaticity diagram………………………………………………………………………………. 6

2.2 Display characterization ...................................................................................................................... 7

2.2.1 LUT model .................................................................................................................................... 8

2.2.2 Masking model .......................................................................................................................... 11

3 Liquid crystal display .............................................................................................................................. 14

4 Equipment and experiments ................................................................................................................... 17

5 Results ...................................................................................................................................................... 18

References ............................................................................................................................................... 28

Appendices

A Instruction for the spectroradiometer

B LCD characteristics and properties

C List of contributions

CHAPTER I

Introduction

The characterization and calibration of display devices we use is an important task. The

Liquid Crystal Displays (LCD) are very popular nowadays and most dominant used

technology in desktop monitors, laptops, smart phones, and many data projection

devices. They have replaced cathode ray tube (CRT) displays, as a flat panel equivalent

which is significant in quality improvement and cost. These LCD’s need an authentic

color management system (CMS) so that the properties of LCD can be defined for

accurate color reproduction. Device characterization is a process of modelling the device

and to measure the properties to get the required image [1]. The goal is to establish a

relationship between digital input values (RGB) and tristimulus values (XYZ). The

characteristics of a good model are that it is fast, requires little data, the calculations are

not complex and it can do a backward mapping from tristimulus (XYZ) values to RGB

color coordinates [2]. So here we have two color spaces. One is the RGB color space which

is device dependent. The other is XYZ color space which is device independent. The XYZ

color space is defined as per the standards of how an observer views a certain color. The

process of device characterization works in two steps. The first step is that color is

produced by giving a set of input values to the device which are the RGB color

coordinates. The characterization of the device requires measurement of the color

produced on it. Using a certain model we can predict that color by finding the XYZ

values, to see how accurately it is viewed. The second step is to do backwards

implementation of the same process where we give the XYZ values and try to reproduce

the desired color on the display device. The relationship mapping between these two

color spaces using a particular model is the whole process of display characterization.

For the purpose of device characterization certain models have been developed. Now

LCD’s could have been characterized by the same models used for the CRT display

characterization but there are a few problems with this approach. LCD’s are more

complicated than CRT’s because of its certain characteristics of channel interaction and

non-constancy of chromaticity and therefore it doesn’t give accurate results. Therefore,

some changes have been made and new models are proposed by researchers to

characterize a LCD device which caters to its specific needs.

In this report we have worked on the LCD characterization of two models. The simple

linear model which utilizes a look up table also called LUT model. The other is known as

Masking model which is relatively new and gives accurate results in case of LCD’s. Both

of the models are discussed in detail in this report.

In the linear model, we estimate the XYZ as a linear combination given by RGB color

primaries. The modeling is done by linearizing the digital input response curve with a

particular nonlinear function. There is a look up table present to associate every XYZ

tristimulus value to RGB and vice versa.

The masking model realizes the color variation in the LCD due to channel interaction and

uses the concept of under color removal to mask inputs from RGB space to RGBCMYK

space. It linearizes the digital inputs and luminance by approximation using a method

known as spline interpolation which is similar to the technique used in linear model [3].

There are several devices which can be used for display characterization. As the

measurements which we need are for the light emitted by the display; we can use the

devices namely, spectroradiometers or colorimeters. The spectroradiometer measures the

spectral power distribution of the light coming from the display. Colorimeters measure

the CIEXYZ tristimulus values.

For the purpose of our experiment we have used a spectroradiometer to measure the data

which is further discussed in our report. To derive the RGB and XYZ relationship and the

nonlinear transfer function of RGB channels we measure a color ramp for each primary

color separately. The values of a particular color channel are measured from zero to its

maximum while the other channels are zero. This spectral data is converted to XYZ values

[4]. Linear model and Masking model are used for processing the data and reproduce the

color using Matlab codes and the results are then compared to check for accuracy by

finding out the color difference.

In retrospect, it is very important in color management that a display device behaves

according to its ICC (international Color Consortium) profile which is a set of

measurements which describes a color input and output of the device and defines a

specific color space.

CHAPTER II

Theory

2.1 Colour coordinates and chromaticity diagram:

There are two basic coordinate systems by which one can produce any desired colour.

The simplest coordinate system is RGB coordinate where R, G, B represents the basis

vectors of colour coordinate system. Here R, G, B represents the amount of red, green and

blue colour. In practice, the addition of two colours produces a very small variety of

different colours, thus three colours are required and the new colour depends on the

amount of the three principle colour that has been mixed. So, RGB system is physically

realizable system.

Figure 2.1: RGB primaries [5]

On the other hand XYZ is another coordinate system where X, Y, Z are the tristimulus

values. They produce another set of basis vectors and the coordinate system defined by

them is different than RGB coordinate. This XYZ coordinate system is completely virtual

and mathematical. XYZ depends on human eye sensitivity. Every linear combination of

XYZ corresponds to unique combination of RGB, means there is one-to-one

correspondence between XYZ and RGB. This XYZ coordinate system is derived by CIE

in 1931 and the derivation of XYZ tristimulus values from RGB physical primaries follows

a linear transformation under certain conditions as follows:

1. The primary colours are chosen in such a way that all real colours have a positive

tristimulus value.

2. Only one of the primaries should be involved to calculate the luminance. This

primary represents the Y coordinate of the XYZ system which represents the

human eye sensitivity curve V(λ) .

3. These new primaries are scaled in such a way that one unit of each primary

together produces equal energy of white.

4. The primaries X and Y are taken in such a way that the line connecting them is

tangent to the spectral locus at 770 nm. As a result the value of the primary Z

becomes negligible at the red end of the spectrum.

The CIE chromaticity diagram represents all the real colours starting from wavelength

380 nm to 780 nm. Though the RGB system specified by the CIE and that for the

display are bit different. The coordinates of the chromaticity diagram x and y are

defined such that,

𝑥 = 𝑋

𝑋 + 𝑌 + 𝑍

𝑦 = 𝑌

𝑋 + 𝑌 + 𝑍

Figure 2.2: CIE 1931 chromaticity diagram [6]

In the chromaticity diagram each value of (x, y) represents a colour. The area under the

triangle joined by the three primaries is called the colour gamut. Any colour within this

triangle can be obtained by the combination of these three primaries. Each display defines

its own colour gamut. All the colours present within the colour gamut can be obtained

for the respective displays but the colours outside the gamut cannot be obtained. [7]

2.2 Display Characterization:

Display characterization is the process to establish a relationship between colorimetric

coordinates and display coordinates. There are two basic models that has been discussed

here to define this relationship namely LUT (look Up Table) model and Masking model.

2.2.1 LUT Model:

LUT or Look Up Table model is basically used to form a curve as a function of ‘R’ vs ‘dr’,

‘G’ vs ‘dg’ and ‘B’ and ‘db’ for some discrete values between 0 to 255. Here R, G and B

defines the irradiance of RED, GREEN and BLUE respectively and dr, dg and db are the

DAC (Digital to Analog Converter) values for RED, GREEN and Blue respectively. Once

the nonlinear curve is plotted then from any R, G or B value one can find out the

corresponding DAC values by simple interpolation. To obtain this curve, first one has to

find out the spectral radiance for RED, GREEN and BLUE colour separately using

Spectroradiometer. Say for Red colour, we have seventeen different combinations of dr

starting from 1/17 to 17/17, while dg and db values are kept fixed at 0 and the readings

are taken starting from the wavelength 380 nm to 780 nm. This curve formed by the

different intensity values as a function of wavelength is called M(λ) curve.

Figure 2.3 : M (λ) curve for red obtained by plotting spectral radiance of red vs.

wavelength for seventeen different values of dr (starting form 1/17 to 17/17), while dg

and db were kept fixed at 0.

Figure 2.4: M (λ) curve for green obtained by plotting spectral radiance of green vs.

wavelength for seventeen different values of dr (starting form 1/17 to 17/17), while dr

and db were kept fixed at 0.

Figure 2.5: M (λ) curve for blue obtained by plotting spectral radiance of blue vs.

wavelength for seventeen different values of dr (starting form 1/17 to 17/17), while dg

and dr were kept fixed at 0.

Each of these M (λ) curves contains seventeen individual curves corresponding to

seventeen different DAC value combination. To find out the tristimulus values

corresponding to each M(λ) curve one can use the following relations.

X = 683 ∫ �̅�(𝜆)𝑀(𝜆)𝑑𝜆780

380 (1)

Y = 683 ∫ �̅�(𝜆)𝑀(𝜆)𝑑𝜆780

380 (2)

Z = 683 ∫ 𝑧̅(𝜆)𝑀(𝜆)𝑑𝜆780

380 (3)

Where, X, Y and Z are called the tristimulus values and �̅�(𝜆), �̅�(𝜆) and 𝑧̅(𝜆) are called

colour matching functions and are provided by CIE.

Figure 2.6: CIE Colour Matching Functions

The relationship between RGB colorimetric values and XYZ tristimulus values is defined

by the following algorithm.

[𝑋𝑌𝑍

] = [

𝑋𝑟,𝑚𝑎𝑥 𝑋𝑔,𝑚𝑎𝑥 𝑋𝑏,𝑚𝑎𝑥

𝑌𝑟,𝑚𝑎𝑥 𝑌𝑔,𝑚𝑎𝑥 𝑌𝑏,𝑚𝑎𝑥

𝑍𝑟,𝑚𝑎𝑥 𝑍𝑔,𝑚𝑎𝑥 𝑍𝑏,𝑚𝑎𝑥

] [

𝑅(𝑑𝑟)𝐺(𝑑𝑔)𝐵(𝑑𝑏)

] + [𝑋𝑌𝑍

]

𝑜𝑓𝑓𝑠𝑒𝑡

(4)

Here, the X, Y, Z tristimulus values are related to the sample. From the above relation we

can find out the R(dr), G(dg) and B(db) values by simple matrix calculations.

[

𝑅(𝑑𝑟)𝐺(𝑑𝑔)𝐵(𝑑𝑏)

] = [

𝑋𝑟,𝑚𝑎𝑥 𝑋𝑔,𝑚𝑎𝑥 𝑋𝑏,𝑚𝑎𝑥

𝑌𝑟,𝑚𝑎𝑥 𝑌𝑔,𝑚𝑎𝑥 𝑌𝑏,𝑚𝑎𝑥

𝑍𝑟,𝑚𝑎𝑥 𝑍𝑔,𝑚𝑎𝑥 𝑍𝑏,𝑚𝑎𝑥

]

−1

[[𝑋𝑌𝑍

] − [𝑋𝑌𝑍

]

𝑜𝑓𝑓𝑠𝑒𝑡

] (5)

The 𝑋𝑟,𝑚𝑎𝑥 , 𝑌𝑟,𝑚𝑎𝑥 and 𝑍𝑟,𝑚𝑎𝑥 these matrix elements are tristimulus values and is defined

by CIE. The [𝑋𝑌𝑍

]

𝑜𝑓𝑓𝑠𝑒𝑡

values can be found out by Spectroradiometer if one puts the DAC

(dr, dg, db) values to (0 0 0) as an input while measuring the spectral radiance. For the

RED now we have seventeen RGB values corresponding to seventeen dr values. Now the

R vs dr curve can be plotted. With this curve we can find out any dr value corresponding

to any R value from the graph.[4] This process can be repeated for BLUE and GREEN

colour. There are different ways for obtaining R(dr), G(dg) and B(db) , the model

described here is one of the possible way.

2.2.2 Masking Model:

If channel interaction is present then the LUT model cannot define the colours properly.

Masking model has been established to reduce the effect of channel interaction. Unlike in

previous model, here we do not only use RGB, but also CMYK, where CMYK represents

Cyan, Magenta, Yellow and shades of Grey.

This model is used to convert any given RGB values to new masked values (RGBCMYK).

To implement this model, it is very important to have the knowledge about the order of

magnitude of dr, dg and db. Suppose we have dr>dg>db.

Figure 2.7: RGB values to masked values conversion

As db has the smallest magnitude, so this much amount will be replaced by Grey. Green

has the second highest magnitude. So it will be replaced by yellow at value dg minus

yellow at value db. Finally Red has the highest magnitude. Thus it will be replaced by

red at value dr minus red at value dg.

𝐼(dr,dg,db) = {𝐼𝑟(𝑑𝑟) − 𝐼𝑟(𝑑𝑔)} + {𝐼𝑦(𝑑𝑔) − 𝐼𝑦(𝑑𝑏)} + 𝐼𝑘(𝑑𝑏) (6)

It shows that masking model is more convenient as channel interaction has been

considered. In LCD display sometimes it may happen that there is some backlight even

when the screen is BLACK. It is necessary to do the black correction when we are

calculating the tristimulus (XYZ) values of single RGBCMYK channel.

𝐼�̂�(𝑑𝑖)= 𝐶𝑖(𝑑𝑖) [

𝑋𝑖,𝑃𝐶𝐴

𝑌𝑖,𝑃𝐶𝐴

𝑍𝑖,𝑃𝐶𝐴

] + I(0,0,0) (7)

Where i = R, G, B, C, M, Y, K

To find the values for X, Y, Z we use a method named Principal Component Analysis

(PCA). In the above equation 𝑋𝑖,𝑃𝐶𝐴 ,𝑌𝑖,𝑃𝐶𝐴 and 𝑍𝑖,𝑃𝐶𝐴 is the first eigenvector of the

correlation matrix calculated from spectral data. If 𝐼𝑖(𝑑𝑖) is the tristimulus values of

digital input di for i= R, G, B, C, M, Y, K, then 𝐶𝑖(𝑑𝑖) can be calculated by the following

equation.

𝐶𝑖(𝑑𝑖) = {𝐼𝑖(𝑑𝑖) − 𝐼(0,0,0)}𝑇 [

𝑋𝑖,𝑃𝐶𝐴

𝑌𝑖,𝑃𝐶𝐴

𝑍𝑖,𝑃𝐶𝐴

] (8)

Where, 𝐼𝑖(𝑑𝑖) is approximated to 𝐼�̂�(𝑑𝑖). To transform from XYZ tristimulus values to RGB

values one can use the following equation.

[𝑋𝑌𝑍

] = {𝐶𝑖(𝑑𝑖) − 𝐶𝑖(𝑑𝑗)} 𝐴𝑖 + {𝐶𝑗(𝑑𝑗) − 𝐶𝑖(𝑑𝑘)} 𝐴𝑗 + 𝐶𝑘(𝑑𝑘)𝐴𝑘 + 𝐼(0,0,0)

(9)

Where 𝑖 stands for primaries (𝑖 = 𝑅, 𝐺, 𝐵) and 𝑗 represents the secondary colours (𝑗 =

𝐶, 𝑀, 𝑌) and finally 𝑘 stands for shades of Grey. Here 𝐴 stand for the

matrix[ 𝑋𝑃𝐶𝐴 𝑌𝑃𝐶𝐴 𝑍𝑃𝐶𝐴 ]. From the above equation it is easy to derive 𝐶𝑖 , 𝐶𝑗 and 𝐶𝑘 by

matrix calculation.

[

𝐶𝑖(𝑑𝑖) − 𝐶𝑖(𝑑𝑗)

𝐶𝑗(𝑑𝑗) − 𝐶𝑖(𝑑𝑘)

𝐶𝑘(𝑑𝑘)𝐴𝑘

] = [𝐴𝑖 𝐴𝑗 𝐴𝑘]−1 ([𝑋𝑌𝑍

] − 𝐼(0,0,0)) (10)

It is possible to find out the value of 𝐶𝑘 and 𝑑𝑘 from equation number (8) and from the

value of 𝑑𝑘, 𝐶𝑖 and 𝐶𝑗 , one can find out the values of 𝑑𝑖 and 𝑑𝑗. [3]

CHAPTER III

Liquid Crystal Displays

Liquid Crystals displays are very commonly used these days in the form of laptop

computers, digital watches, flat screen television sets and many other devices. They have

replaced CRT displays due to their low cost, compactness and low power consumption

which is roughly in terms of one half what CRT requires. They have better contrast and

better brightness. Here we will discuss the working principle of an LCD and will go

behind the science of liquid crystals and how they produce colors.

A liquid crystal display is an arrangement of tiny segments which are known as pixels.

These pixels can be manipulated to present color information. A basic LCD has two glass

plates with transparent electrodes inside their surfaces and a liquid crystal material

sandwiched between them. These liquid crystals are substances in a state between solid

and liquid. The liquid state refers to the outer appearance of the substance and the solid

state refers to the orientation of the molecules inside the substance. These molecules tend

to arrange themselves in particular ways which enables them to flow as a liquid and react

to electric current. This particular way in which all molecules are in the same direction is

known as nematic phase.[8] These liquid crystal can control the pixels and turn them on

and off by using an property of light known as polarization, which means to filter the

light so that it can travel in only one direction or plane. This is possible as LCD cannot

produce light on its own and requires an external illumination to create visual effect. So

the basic structure of the LCD depends on the applied electric field which enables the

liquid crystals to transmit and polarize the light coming from an external source which

in turn controls the pixels.

Here we will discuss the construction and working of the most commonly used type of

LCD, the twisted nematic liquid crystal. The structure of these can be made in such a way

that that the difference between extraordinary and ordinary refractive indexes is ideal for

LCD usage.[9]

The construction of the device consists of a layer of uniaxial liquid crystal sandwiched

between two glass substrates. Here, it is essential to define the director which is “the

average direction of alignment over a local region containing many molecules”[9]. In

nematic liquid crystal, the director means the local optical axis. A layer of polyimide is

deposited on the glass surfaces and is rubbed it in a fixed direction which in turn aligns

the liquid crystals in the same direction. If the rubbing directions on both the surfaces is

orthogonal, the director on the first substrate gets rotated by 90 degrees with respect to

the other substrate. Hence the director is forced to induce a twist of 90 degrees of the

liquid crystal between the two glass surfaces. Now the liquid crystal is placed between

two polarizers with the transmission axes aligned in the same direction as the rubbing of

polyimide surface. Which means that the polarizer at the input end and the output end

are now crossed.

Figure 3.1: Working of twisted nematic LCD in normally white mode. The arrows show

the transmission axis of polarizers. [9]

When dealing with nematic liquid crystals, there are two modes of operation for this

device. These depend on the orientation of the polarizers. If the polarizer at the output

end is parallel to the one at the input end, it is called normally black mode and if the

output polarizer is perpendicular to the input polarizer it is called normally white mode.

In normally white mode, which the more commonly used, if there no is voltage applied,

light enters the liquid crystal through the input polarizer following the same twisted

orientation as the liquid crystal molecules which changes its polarization by 90 degrees.

So when it reaches the output polarizer it is parallel to its transmission axis and hence

light passes. This is called the field off state. But when voltage is applied the liquid

crystals are affected by the electric field which is perpendicular to the liquid crystal and

molecules align along with it and because of this the polarization of light doesn’t change

and when it reaches the output polarizer it is perpendicular to its transmission axis. Thus

light cannot pass and it is blocked. This is called the field on state. This is shown in figure

3.1.

In normally black mode, which is less common. The operation is completely opposite.

When there is no applied voltage, the light at the output polarizer is perpendicular to its

transmission axis and light is blocked. But if there is applied voltage the molecules of the

liquid crystal get aligned with the electric field and when light reaches the output

polarizer it is polarized by 90 degrees and hence parallel to its transmission axis and light

passes[10].

The LCD construction also requires a backlight, and two electrodes to be deposited on

the glass surfaces. A LCD is divided into pixels and sub pixels to produce color. In that

case we have color filters for red, green and blue. A thin film transistor is located at each

pixels to turn the pixel voltage on and off. This is shown in figure 3.2.

Figure 3.2: Compenents of a twisted nematic LCD.[9]

An LCD color gamut describes a specific range of colors produced by the device which

are identifiable by a human eye. A color gamut is represented by a xy chromaticity

diagram which is generated by numerical values of chromaticity. The area is shaped like

a ‘U’ represents the colors visible to humans. The color gamut of a LCD is defined by a

triangle indicating the colors it can reproduce [11]. An example is shown in the figure 3.3

There are several drawbacks also when LCDs are concerned which are that they require

an additional light source, they have a limited temperature range for operation, poor

visibility in low lighting and slow speed. The most common defect is weak pixels. Due to

these disadvantages some issue arise during characterization of an LCD. For instance the

different viewing angle in LCDs lead to different color appearance. This is due to the

property of birefringence exhibited by liquid crystals [7]. Other issues can arise due to

flare from the screen, the spatial dependence which refers to effect a color displayed on

an areas has on a color displayed in another area [12]. Also we have to assume that a pixel

is not affected by neighboring pixels and there is enough time is given for the monitor to

stabilize and reach a steady state which is also known as temporal stability of the device.

Figure 3.3 : Color gamut of an LCD TV based on CIE1931 color system with D65 as

illuminant for white point [13]

CHAPTER IV

Equipment and experiments

For the purpose of our experiment we have used a spectroradiomater with a Dell monitor

display to do the characterization measurement. The steps to operate and take the

measurements are included in Appendix A. Initially we used the laptop screen for Dell

Latitude D620 for our measurments. But due to several issues with the brightness and the

displaying of color blue we had to change the display. The laptop display characteristics

and issues are mentioned in Appendix B. The final measurements have been done with

a Dell monitor display. After the lab measurements we have used those measurements

to develop the matlab code to apply the linear model and the masking model. We have

converted the final results to CIELAB color system to find the color difference to check

the accuracy of both the models using Matlab.

CHAPTER V

Results

Two models were used to predict the DAC values that are needed to reproduce a color

on a monitor from a defined XYZ tristimulus values of a sample color. (Macbeth

colorchecker, Fig.5.1, is our sample)

Figure 5.1 Macbeth ColorChecker, the numbers will be used later on to refer to each

patch

5.1 Linear Model:

One model assumes that the CIE tristimulus values of the monitor displayed color are a

linear combination of the tristimulus values of three independent monitor channels (Red,

Green, and Blue). i.e. X(displayed color)=Xred+Xgreen+Xblue; Y(displayed

color)=Yred+Ygreen+Yblue; Z(displayed color)=Zred+Zgreen+Zblue. That’s why I will be

referring to this model as the linear model from now on.

The monitor’s channels tristimulus values (Xred,Yred,Zred,Xgreen,…) are of course

dependent on the spectral radiance of the channel. Now in this linear model, the channel’s

spectral radiance is assumed to have exactly the same profile regardless of the level to

which the channel is activated (whether it is 10 or 255 level of an 8-bit channel for

example). This means that when a particular channel is activated to different DAC values,

the spectral radiance profile will stay the same, only multiplied by a factor, a scalar, which

will be denoted by R for red, G for green, and B for blue. This assumption is known as

channel constancy and these scalars (RGB) follow a nonlinear relationship with the DAC

values as shown in figure 5.2. (This assumption originated from the operation of CRT

displays where the relative spectral radiance of the phosphor emission is constant.)

We applied this linear model to the sample shown in figure one. The XYZ values of the

numbered patches are fed to a simple program which calculates the RGB scalars at first

(based on equation 5) and then uses spline interpolation to calculate the DAC values. The

spline interpolation relies on the different values of R,G, and B (18 values each) that were

calculated from the spectral radiance of each channel.

Figure 5.2: R,G,B vs the DAC values for each channel

A visual comparison was made between the monitor generated colors and the Macbeth

color chart for which the agreement was good when viewed from a certain angle. (In

principle the agreement shall occur when the line of sight is normal to the screen given

this is the way the spectral radiance data was collected from the display for different

ramps.)

To quantify whether a match is obtained, the color difference E* is calculated using the

CIE L*a*b* coordinates. Figure 5.3, shows the color difference for each patch of the

Macbeth sample.

Figure 5.3: Color difference vs sample's color patch number

It’s to be noted that the spline interpolation led to a large error in predicting the DAC

values to the nature of the obtained data set. So to be able to use the spline interpolation,

a simple trick was devised here. Rather than interpolating the DAC values from

calculated R,G, and B values of the measured spectra at different ramps, a very big set

(2550 values) of R,G, and B values were interpolated first from the measured R,G, and B

values and then the DAC values for our test sample was interpolated from this new

bigger set. This trick provided the results of Figure 5.3. As seen in the figure, the color

difference is between 2 and 4 for most of the patches in the tested sample (average E* is

3.39). The reason for this difference could be related to four factors:

1. Viewing angle (we could have obtained better match if the spectroradiometer was

adjusted at the optimum position, i.e. to resemble the one used when the data was

collected)

2. Human error (a mishap could have seeped into the code, we had no time for

second trials and double checking)

3. The display is producing slightly different colors at different times of operation

4. The Imperfection of our model. Let us remember that we are:

i. Interpolating data based on 19 measurements

ii. Assuming channel independency and chromaticity constancy. Figure 4 and

5, shows clearly that these assumptions are not entirely true.

Figure 5.4 : Diplay’s luminance difference in Cd/m^2 between the channels considered

separately and combined. Blue curve considers the RGB channels, green curve assumes

a cyan, magenta, and yellow channel

In figure 5.4 above, the fact that the luminance of the display when all the channels were

set to a certain value is smaller than the luminance of the display when calculated from

the addition of the luminances of each channel taken separately tells us that while only

one channel is activated, the others are still contributing to the display’s luminance. This

means that the channel’s are not totally independent, i.e. Red and green are present when

blue is activated and the same for the other channels.

The green curve in figure four shows us that this rough quantitative measure of channel

dependency is minimized if cyan, magenta, and yellow are considered as channels, i.e. a

combination of channels is considered here as independent channels. I did this plot to

suggest that the linear model can be optimized by using this combination of channels.

Simple adjustment has to be done to the linear model as stated in equation five. A factor

of half need to be added to the matrix and red, green, and blue will have to be replaced

by cyan, magenta, and yellow. This method has not been investigated by us at this stage.

Figure 5.5 below, shows that the chromaticity coordinates of the primary and secondary

colors is actually changing for the channels when activated at different levels, different

DAC values.

Figure 5.5: Chromaticity shift of the primary and secondary colors. The one on the Left,

the offset is not included, while the one on the Right, offset is included

5.2 Masking Model :

In this model the CIE tristimulus values of the monitor displayed color is a linear

combination of the tristimulus values of a grey channel, a secondary channel, and a

primary channel. These channels do not exist in reality, they are a mathematical

abstraction to take into account the channel interaction. This technique is inspired by the

under color removal technique that is used in printing and is developed by Tamura,

Tsumura and Miyake [3].

In our experiment the tristimulus values of each of these new channels are considered to

lie along a line when plotted in an XYZ Cartesian coordinate system. This is like the

spectral radiance profile assumption that has been made in the linear model. The

direction of this line is given by the direction of the first principal component as shown

in figure 5.6 below for the magenta channel.

Figure 5.6: XYZ in magenta channel, the solid line gives the direction of the first

principal component

When the direction of the first principal component is known the XYZ of the color

produced by any of the masking model channels can be calculated by multiplying that

direction by a scalar. The scalar for each of these channels (follow nonlinear behavior) is

denoted by Ci in equations 8,9, and 10 and is shown in figure 5.7 for all the masking

model channels.

Figure 5.7: Magnitude of the “projected XYZ vector of the primary and secondary

channels on the first principal component” at different DAC values, the C values.

We applied this model to the same sample shown in figure 5.1. The XYZ values of the

numbered patches are fed to a simple program which

1.Calculates the DAC of the red, green and blue channels from the linear model, just to

figure out the primary and secondary channel that is to be used for the masking model

2.The right hand side of equation 10 is then calculated, it uses the first principal

component directions explained above for each channel. This will give us the “C scalar”

for the grey channel.

3.The DAC values of the grey channel is interpolated first from the “C” scalar and this

DAC value is the same as that of the least active channel among the red, blue or green.

(We know which in particular from step 1).

4.The “C” scalar of the secondary and primary channels is calculated working from

bottom to top in equation 10.

5.The DAC values of the masking model primary and secondary channel is then

interpolated. From which the DAC values of the most and second most active red, green,

or blue channel is calculated.

This awkward description is given in a rather neat form in equation 10.

Using this model we obtained a better display of the Macbeth sample on our display. CIE

L*a*b* is again used to calculate the color difference and the result is shown in figure 5.8.

Figure 5.8: color difference vs average DAC value of the color checker patches

Using this model the average color difference was about 1.8 which is an improvement

over the linear model. In figure 5.8, the blue solid line shows the channel interaction that

was described in figure 5.4 to see how the channel interaction contributes to color

difference in the masking model.

To test the masking model in another way, random colors were generated on the display.

Their spectral radiance was measured and the DAC values were obtained. After that the

DAC values were used to generate the colors again to check whether we will obtain a

match. In figure 5.9, we show the obtained result.

Figure 5.9: Left: Randomly generated color on the display Right: color obtained from

calculated DAC values

REFERENCES

References:

[1] Brainard, D. H., Pelli, D. G. and Robson, T. “Display Characterization. Encyclopedia

of Imaging Science and Technology”, 2002.

[2] Bastani, B., Cressman, B. and Funt, B. (2005), Calibrated color mapping between

LCD and CRT displays: A case study. Color Res. Appl., 30: 438–447.

[3] Tamura, N., Tsumura, N. and Miyake, Y. “Masking model for accurate colorimetric

characterization of LCD”, Journal of the SID, Volume 11, Issue 2, pages 333–339, June

2003.

[4] Laamanen, H. "Spectral color and spectral image analysis," University of Joensuu,

Joensuu, 2007.

[5] (n.d.). Retrieved March 24, 2016,

https://en.wikipedia.org/wiki/File:AdditiveColor.svg

[6] (n.d.). Retrieved March 24, 2016, https://en.wikipedia.org/wiki/RGB_color_space

[7] Gibson, J., Fairchild M. D. “Colorimetric characterization of three computer displays

(LCD and CRT)”. Munsell Color Science Laboratory Technical Report, 2000.

[8] LCDs (liquid crystal displays). (n.d.). Retrieved March 13, 2016,

http://www.explainthatstuff.com/lcdtv.html

[9]Kenyon, I., “The Light Fantastic: A Modern Introduction to Classical and Quantum

Optics”, Oxford University Press, Inc., New York, NY, USA, 2008.

[10]Liquid Crystals: A Simple View on a Complex Matter. (n.d.). Retrieved March 24,

2016, http://www.personal.kent.edu/~bisenyuk/liquidcrystals/applic1.html

[11] The Ability to Display Color Correctly Is Vital: Understanding the Color Gamut of

an LCD Monitor | EIZO. (n.d.). Retrieved March 13, 2016,

http://www.eizo.com/library/basics/lcd_monitor_color_gamut/

[12] Khisa, S. “Characterization and calibration of a LCD display based on a limited set

of color samples”, Université Jean Monnet, 2010.

[13]Retrieved March 24, 2016, from https://en.wikipedia.org/wiki/Rec._709

[14] Specifications. (n.d.). Retrieved March 13, 2016,

http://www.solano.edu/technology/data/D620/specs.htm

[15] Review Dell Latitude D620 Notebook. (n.d.). Retrieved March 13, 2016,

http://www.notebookcheck.net/Review-Dell-Latitude-D620-Notebook.3491.0.html

APPENDIX A

Instructions for the spectroradiometer

Equipment Used:

-Konica Minolta CS2000

spectroradiometer

-Laptop display

-Dell computer display

-Data cable

-Usb key

-Tripod stand

Steps:

Adjusting the spectroradiometer on the tripod stand properly. The tripod stand is to be

aligned properly.

Placing the laptop in front of the spectroradiometer. The screen of the laptop is to be in

an upright position; perpendicular to the viewing direction.

Making the necessary connections between the camera and the computer. Switching on

the camera and the computer and insert the usb key.

Writing a code on matlab to generate a color and focus the spectroradiometer spot on it.

The spot size can be change with regard to the distance between the

spectroradiometer and the display.

Opening the software “CS-SIOw” and choosing “normal mode”.

Then we have to click “instrument” and then click “connect” to make the connection

between device and display

Then “file” “template” “load new template” to load the required template.

Giving some time for the laptop screen to stabilize. In our case it will be approximately

40 minutes.

Turning off the light and then taking the required readings

Saving the readings as text file by going to “file” “save selection as text” so that it

can used later for processing

We also have repeated all the measurements with a Dell monitor for better results.

APPENDIX B

LCD characterisitics and properties

LCD specifications and characteristics (Latitude D620):

In the lab we used a laptop screen Dell Latitude D620 for the purpose of our experiment.

Here we list the several display properties of the LCD display [14].

Viewing size: 14.1 inches

Height x Width x Depth: 1.26 x 13.27 x 9.37 inches

Type: Active matrix TFT Color LCD

Resolution: WXGA (1280 x 800) or WXGA+ (1440 x 900) display

Weight: 2.3 kg

Aspect ratio: 16:10

Maximum Brightness: 142.7 cd/square meter

Minimum Brightness: 0.6 cd/ square meter

Pixel Pitch: WXGA (0.2588) WXGA+ (0.1971)

Power Consumption: WXGA (5 W) WXGA+ (5.5 W)

Contrast: 238:1

116.2

cd/m²

124.7

cd/m²

102.3

cd/m²

130.2

cd/m²

142.7

cd/m²

127.5

cd/m²

131.5

cd/m²

128.2

cd/m²

120.5

cd/m²

Figure 4: Distribution of Brightness [15]

The D620 has a built-in ambient light sensor built-in that will control the screen

brightness when on battery. For example, in a dark room the screen will dim as less

brightness is needed to see the screen, but in bright light the screen will increase in

brightness so it is easier to see. When on power the screen will automatically become

brighter. You can override the ambient light sensor to control brightness which is done

by the function key and arrow keys.

Due to several disadvantages of using this display we have again performed our

experiment with a Dell computer display where we have adjusted the color settings for

brightness ourselves.

The color diagram [15] depicts an ideal red and green color curve, but also the usual

lowered blue color curve. So, red colors dominate.

While operating the laptop, from the control panel and choosing display properties you can

change the following parameters as shown in the figures below:

APPENDIX C

List of contributions

Yousuf Hemani:

I have worked on the abstract, introduction, working and construction of the lcd display, viewing

defects, lcd gamut, equipment and experiment, appendix A and B. I have participated in the Lab

measurements and the performing of the experiment to get results. I have helped in compiling

the report and I have provided feedback to others whenever it was needed.

Tanmay Mondal:

I have made the working plan, the whole theory part including colour coordinates and

chromaticity diagram, display characterization, LUT model, masking model. At each step I have

consulted and discussed everything with my supervisor and group members. I have also

participated in lab measurements and provided feedbacks to others. Finally I have done the

whole compilation process of the report with the help of Yosuf Hemani.

Mohammed Al Lakki:

I enjoyed working on this project. I have learned a lot about color science through it. Though a

lot of time was spent in characterizing an improper display, however this taught me about how

unexpected delays can occur in experimental work and that it is a factor that has to be included

in any preparation. I spent more than the expected time for this report, and I have utilized the

computational skills that I have learned in the matlab course that I have taken. I have participated

in collecting the data, wrote the matlab codes, chapter V and provided feedback to others before

chapter five.