cinema screen uniformity measurement method

cinema screen uniformity measurement method
CINEMA SCREEN UNIFORMITY MEASUREMENT METHOD
by
Erik Heath Arend
A Document Submitted to the Faculty of the
COLLEGE OF OPTICAL SCIENCES
In Partial Fulfillment of the Requirements
For the Degree of
MASTER OF SCIENCE
In the Graduate College
THE UNIVERSITY OF ARIZONA
2015
2
TABLE OF CONTENTS
LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3
ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4
CHAPTER 1 Introduction
. . . . . . . . . . . . . . . . . . . . . . . . . . .
5
CHAPTER 2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6
CHAPTER 3 Overview of Method . . . . . . . . . . . . . . . . . . . . . . .
8
CHAPTER 4 Measurement Setup . .
4.1 BRDF Goniometer . . . . . . .
4.2 Projector & Camera . . . . . .
4.3 Reference Measurement System
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
10
10
12
17
CHAPTER 5 Data Processing Algorithm . . . . . . . . . . . . . . . . . . . 18
5.1 Processing Method for Image Based Method . . . . . . . . . . . . . . 18
5.2 Processing Method For Reference Systems . . . . . . . . . . . . . . . 21
CHAPTER 6 Results
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
CHAPTER 7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3
LIST OF FIGURES
4.1
4.2
4.3
4.4
4.5
4.6
5.1
5.2
BRDF Goniometer . . . . . . . . . . . . . . . .
Example Gain Curve . . . . . . . . . . . . . . .
Projection and Capture Setup . . . . . . . . . .
Photo of Projector and Camera Setup . . . . . .
Photo of Lab Setup. Projector is on the left and
the right . . . . . . . . . . . . . . . . . . . . . .
Photo of Vacuum Wall . . . . . . . . . . . . . .
. . . . .
. . . . .
. . . . .
. . . . .
Vacuum
. . . . .
. . . . .
. . .
. . .
. . .
. . .
Wall
. . .
. . .
. . .
. . .
. . .
. . .
is on
. . .
. . .
.
.
.
.
12
13
14
15
. 16
. 17
5.3
5.4
5.5
5.6
Geometry of the Projected and Viewer Rays . . . . . .
Map of Angles from the Specular Reflection for each
image. . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Gain Map of Screen Sample . . . . . . . . . . . . . . .
Data Processing Sequence . . . . . . . . . . . . . . . .
Comparison to Reference System . . . . . . . . . . . .
. . .
pixel
. . .
. . .
. . .
. . .
. . .
. . . .
in the
. . . .
. . . .
. . . .
. . . .
. . . .
.
.
.
.
.
20
21
22
22
23
6.1
6.2
6.3
6.4
Example of Localized Defect . . . . . . .
Average Horizontal and Vertical Profile .
Example of Good Down Web Uniformity
Example of a Virtual Screen . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
25
26
27
28
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. 19
4
ABSTRACT
Polarization preserving cinema screens often suffer from non-uniformities within
the viewing area. These non-uniformities appear bright or dark relative to the
surrounding area and can be very visible to the viewers. The current methods to
assess these non-uniformities are often visual and not quantitative. The following
presents an image-based process for quantifying the uniformity of cinema screens.
By incorporating a known gain profile of the screen sample, this method is able to
remove the viewing patterns associated with view angle and return a flat field image
showing the non-uniformities. This image-based approach additionally provides
higher resolution data in less time than laser-scanning systems built for the same
purpose.
5
CHAPTER 1
Introduction
The polarization preserving screens (silver screens) used in today’s 3D cinema environments and screens in general, can suffer from various forms of non-uniformities.
These non-uniformities typically show up as brightness changes across the screen. As
the technology to fabricate cinema screens continues to progress, the overall screen
performance increases and these non-uniformities can become more apparent to the
viewer. Thus, the measurement techniques used to quantify the screen performance
must also evolve. Historically, the inspection methodologies have been visual or
taking basic photographs. These methods can be useful when the properties of the
screen are changing slowly. However, if the performance changes rapidly, such as a
sudden change in brightness at a seam in the screen, these methods may not be able
to accurately quantify the brightness change and therefore lead to improper screen
construction.
As such, a more rigorous inspection method must be used to identify and quantify the level of non-uniformities that may exist in the manufacturing process. The
following describes an image-based method, developed internally at RealD, which
can be used to measure the uniformity of reflective screen material. This measurement process is also capable of higher resolution data in a shorter amount of time
compared to laser-scanning systems. The system works by capturing luminance images of the screen sample and incorporating information about the screen properties
to generate a map of the screen uniformity. Very few assumptions are made about
the screen material being measured and therefore the performance of the screen can
be properly assessed in a stable and repeatable way.
6
CHAPTER 2
Background
The technology used to manufacture polarization preserving cinema screens used for
3D projection has gone relatively unchanged since polarization based 3D systems
began to grow and dominate the 3D cinema market starting in 2005. The basic
process today is to hang a large piece of vinyl on a frame and use automated equipment to spray a metal-based reflective pigment onto the vinyl. Although it is the
standard screen manufacturing method today, this process results in screens that
have poor stereo contrast ratio, poor total integrated scatter (TIS), visible texture,
and/or ”hot spot” on screen.
A new screen technology in development by RealD aims to solve the deficiencies
of the standard silver screens. As described by Coleman and Sharp (2013), this
new method uses a replication process that directly forms the surface of the screen.
Directly forming the surface in this manner allows for a smooth and continuous
geometry, and thus allows the screen to have high TIS (90%), stereo contrast ratio
greater than 250:1, no visible texture, and no hot spot. All of these properties
combine to make a screen this reflects nearly twice the light and more than three
times the stereo contrast compared to the current silver screens, thus allowing the
screen to seemingly disappear from the viewing experience.
The basis of the screen manufacturing is a roll-to-roll replication process. The
process starts with a large roll of a plastic substrate. Using a cylindrical tool, a UV
curable resin is cast onto one side of the substrate. The cylindrical tool contains
the negative version of the desired microstructure. In this way, the desired surface
gets cast onto the base substrate in a continuous process. The roll of film is then
aluminized using a roll-to-roll evaporative coating. Additional protective coatings
7
are also applied to the film. The film first undergoes a pre-slit step to narrow width.
The film, still in single roll, needs to be assembled into a screen. A specialized
seaming table cuts strips of film off of the main roll, slits the film to the final width,
and seams adjacent panels together. This process repeats until the screen has been
assembled to the desired size.
The replication tool is the basis of the screen performance. Any irregularities
during the fabrication of the tool will be replicated in the film and therefore will
show up in the final screen as changes in gain (brightness) visible across the screen
as a non-uniformity. These types of uniformity concerns are very different than
the random non-uniformities that are introduced by the painting process of typical
silver screens. The shape of these irregularities can vary from small-localized defects
to a gradual change in the gain profile that extends across the entire tool. These
gradual (or sometimes not very gradual) changes can occur in both the down web
(machine direction) and cross web directions. Due to the roll to roll casting process,
these irregularities will also repeat at each revolution of the tool. In this case, the
small localized non-uniformities will show up as spots across the entire screen. A
cross web slope in the gain profile, ”wedge shape”, to the uniformity profile can
be the most problematic for screen construction. When visually inspecting a single
piece of film, the film may appear uniform due to its gradual change in brightness.
It is only when two pieces of film are seamed together that the ramp in brightness
can be seen due to a large brightness jump at a seam between two panels. The
manufacturing steps involved to get to seamed screen can be lengthy and costly.
Therefore, it is beneficial to be able to locate and quantify the uniformity at the
earliest stage possible.
8
CHAPTER 3
Overview of Method
When projecting light onto a screen or sample of screen, the brightness, denoted
B(x, y) and units of lm/m2 sr, seen by the viewer can be described by the combination of three factors. The first is the illumination profile coming from the
projector, denoted as P (x, y). The second is the gain profile of the screen, denoted
as G(θi , φi , θv , φv , x, y). This term is a ratio of the bidirectional reflectance distribution function (BRDF), denoted fr , of the material and a reference sample and is
therefore unit less. It is dependent on the incident angles (θi and φi ), the viewing
angles (θv and φv ), and the spatial location on the screen(x, y). It is assumed in this
situation that the change in the BRDF is dependent on the spatial location and not
the incident or view angles. Thus making the following assumption:
dfr dfr
dfr dfr
,
,
dx dy
dφi,v dθi,v
(3.1)
This assumption is valid for screens at near normal incidence and view angles, and
thus at the peak of the BRDF curve. Making this assumption then allows for a
spatial uniformity term to be separated out from the gain term, denoted as U (x, y).
For an ideal screen, the uniformity function would be unity and therefore does
not contribute to the observed brightness. These three factors combine according
to equation 3.2. Each of these factors is describing the brightness or change in
brightness in terms of the spatial location, in x and y coordinates, on the screen
surface. This also assumes the viewer is fixed. Each of these functions will change
as the location, relative to the screen surface, of both the projector and viewer, are
changed.
9
B(x, y) = P (x, y)G(θi , φi , θv , φv )U (x, y)
(3.2)
Solving for the Uniformity function then results in equation 3.3.
U (x, y) =
B(x, y)
G(x, y)P (x, y)
(3.3)
Therefore, calculating the uniformity of the screen requires knowledge of the
brightness seen by the viewer, the gain profile, and the projector illumination profile.
This method describes in detail the process needed to determine each of these
factors. The overall brightness function, B(x, y) is captured using a calibrated imaging photometer. The camera, in this case, represents the viewer, and thus the spatial
cooridinates are represented by pixels in the image. The gain profile, G(θi , φi , θv , φv )
is measured on a small representative sample using a goniometer based BRDF measurement tool. Using this information, the gain profile is then expanded into a gain
map based upon the projector location and the viewing location. The projector profile, P (x, y) in units of lm/m2 sr, is calculated by using a flat white reference sample
in place of the screen sample. For this term, it is assumed that the reference sample
is uniform (Uref (x, y) = 1) and has a known gain profile close to 1 (Gpaper (θ, φ) ≈ 1).
With all of the raw data collected, each data set is then processed in such a way to
ensure they are all in the same coordinate space and a pixel by pixel calculation can
be performed resulting in the a final map of the screen uniformity.
10
CHAPTER 4
Measurement Setup
In order to assess the uniformity of the base screen material, two sets of measurements need to be taken of the film. The first data set is a localized gain curve. The
second is an image of a large screen sample when illuminated by a cinema projector.
4.1 BRDF Goniometer
The localized gain curve is measured with a goniometer based system measures the
BRDF of the sample. The BRDF of a sample is described by the ratio of the reflected
radiance to the incident irradiance. The BRDF can also be written in terms of the
power and geometry factors. These relationships are given in equation 4.1.
fr =
Lr
Pr /ΩA cos θ
=
Ei
Pi /A
(4.1)
This system resembles the bidirectional scatter distribution function scatterometer described by Stover (2012). A green laser is incident on a stationary sample. A
swing arm with a detector rotates around the sample and measures the amount of
light at each angle. The system is capable of scanning an angular range of θ = −84◦
to θ = 84◦ . Because the detector blocks the incident light, angles between θ = −6◦
and θ = 6◦ are not collected. A diagram of this setup is shown in Figure 4.1. This
system only scans in a single arc, and therefore only captures a small amount of the
entire BRDF function at a time. The sample must be rotated around the optical
axis (defined by the axis of the laser) and remeasured in order to obtain a full hemisphere of data. For the purposes of this testing, data is only collected at φ = 0◦
and φ = 90◦ . In order to measure the peak of the gain curve, which occurs at the
11
specular reflection angle, the sample is rotated to 5◦ around the Ω axis. This places
the specular angle at θ = −10◦ (in machine coordinates). A diagram of this setup
is shown in Figure 4.1.
The gain curve of the material is not a direct measurement. As mentioned before,
the gain of the material is relative to a reference screen that scatters light equally in
all directions and appears equally bright from all directions. This screen would have
a radiance of L = 1/π(sr−1 ) and is commonly referred to as a Lambertian Diffuser.
For each gain curve, measurements of two samples must be collected. The first is
the screen sample itself, including both parallel and crossed polarizers. The second
measurement is a piece of the reference screen. For this testing, Spectralonr is used
as the lambertian reference sample. The equation for screen gain in a 3D viewing
environment is given by 4.2. The BRDF of the screen in the crossed polarization
state is not included in this equation because it is blocked by the eyewear worn by
the viewer and thus does not contribute to observed brightness.
G(θ) =
BRDFScreenk (θ)
BRDFLambertiank (θ) + BRDFLambertian⊥ (θ)
(4.2)
All the measurements included in this equation make use of the same geometry,
solid angle, detector area, sample area, etc. Therefore the geometry terms in the
BRDF equation drop out and the result is the reflected power measured by the
detector as shown in equation 4.3. The output product is a gain vs angle curve
similar to the one shown in Figure 4.2.
G(θ) =
PScreenk (θ)
PLambertiank (θ) + PLambertian⊥ (θ)
(4.3)
Additionally, by placing the appropriate polarizers in the incident beam and at
the detector, the system is also capable of measuring contrast ratio as a function of
angle. Both linear and circular polarization contrast ratios can be measured. This
is achieved by measuring the sample first with parallel polarizers and then again
12
Figure 4.1: BRDF Goniometer
with crossed polarizers. The contrast ratio of the screen is then given in equation
4.4. Again, since these measurements use the same geometry, these terms can be
removed from the calculation and only the reflected powers for each sample are
required.
SCR(θ) =
BRDFk (θ)
Pk (θ)
=
BRDF⊥ (θ)
P⊥ (θ)
(4.4)
4.2 Projector & Camera
The second measurement system captures data from a large piece of screen material.
The large screen sample is placed vertically on a vacuum wall. This holds the sample
flat with no deformations at the edges. Any mechanical deformations in the film will
show up as a non-unifiormity in the final result and therefore needs to be removed
from the setup. The vacuum wall is capable of holding a sample 38 inches wide and
90 inches tall. This corresponds to the full width of the raw film and 2-3 revolutions
of the casting tool.
13
1.5
Gain
1
0.5
-80
-60
-40
-20
0
20
40
60
80
Angle (deg)
Figure 4.2: Example Gain Curve
A cinema projector provides the illumination for this measurement setup. The
projection lens is centered horizontally and vertically to the center of the vacuum
wall. The projector is placed 11.25m away from the vacuum wall. At this distance,
the vertical dimension of the screen sample subtends an angle of approximately ±10◦ .
A full white test pattern from the projector is displayed on the screen sample. A
slight amount of defocus is added to provide a slight blur of the pixels at the screen.
This helps to eliminate moiré between the projector pixels and the camera pixels.
Furthermore, the projected image is zoomed down and cropped so that only the
screen sample is being illuminated. This minimizes the amount of stray light in the
room that can reflect back onto the screen. Both the projector and the vacuum wall
are fixed in place and do not move. This provides a stable and consistent platform to
measure the geometry of the setup. The geometry plays a key role in the processing
of the data and must be known very well. Additionally, the cinema projector is
a good platform to use for illumination because the light output is very stable.
The high powered xenon lamps provide a very bright image, on the order of 40
footLamberts (fL) in this setup, that is stable to 1% RMS (Donohue and Fernandes,
14
2004) over the span of several hours. This allows the testing to be completed over
the course of 1-2 hours without the need to recalibrate the light levels. However,
over longer periods of time, the xenon lamps do degrade. A typical lamp will last
between 1000-2000 hours. For this reason, a new calibration image is required at
the start of testing each day.
The third component to this system is a camera. The camera is mounted on a
tripod and can be placed anywhere in the room. This allows the freedom to capture
data at a variety of angles and distances if required, as long as the assumptions in
equation 3.1 are still valid. For our testing, the most common camera location is
right at the projector. This gives the best location to replicate the uniformity of the
screen on axis and at ‘infinity‘. When placed at the projector, the camera resolution
and field of view equates to a resolution of 5mm per pixel at the screen sample. A
diagram of this setup with the camera at a viewing angle of 30◦ is shown in Figure
4.3.
Figure 4.3: Projection and Capture Setup
15
The camera used for these images is an off the shelf calibrated imaging photometer from Westboro Photonics. The camera sensor is a black and white, 2MP
sensor with a photopic filter placed on top. The filter shapes the spectrum in order
to provide luminance data. This allows the camera to return a luminance value for
each pixel instead of digital counts. The camera also has a high dynamic range,
ranging from 0.06cd/m2 to 100, 000cd/m2 , within a single measurement. This is accomplished by electronic bracketing within the software. The system takes images
over a range of exposures and then on a pixel-by-pixel basis, chooses the appropriate
image from the different exposures to pull from in the final luminance image. The
exact details of how this decision is made are proprietary within the manufacturers
software. The camera captures images at exposures ranging from 2 miliseconds to
8 seconds. When capturing the image, the software also captures 10 images at each
exposure level and then averages those prior to creating the luminance image. This
allows any small temporal fluctuations in the projector illumination to be averaged
out and produces far more reliable data. In the end, approximately 100 images are
used to create the final luminance image.
Figure 4.4: Photo of Projector and Camera Setup
16
Figure 4.5: Photo of Lab Setup. Projector is on the left and Vacuum Wall is on the
right
When capturing the images of the screen sample, at least two luminance images
are need for each camera position. The first is a calibration image used to collect
information about the physical setup and the projector. Similar to the gain calculation, I use a uniform white diffuser for this. The sample material must cover the
same area on the vacuum wall as the screen sample. This large size of diffuser means
that a Spectralon sample is not feasible and bright white copy paper is used in its
place. Individual pieces of the white paper are placed on the vacuum wall until the
entire area is full. This measurement of white paper provides information about the
projector brightness and the projector illumination profile. It should be noted that
the gain profile of this white paper is also measured using the BRDF goniometer.
This allows for a secondary reference back to the primary Spectralon reference. This
information is required in the data processing steps to remove the contributions of
the projector to the uniformity of the screen sample. The second image is of the
screen sample itself. If data from multiple viewing angles are required, a tripod is
setup at each location in advance. This allows the samples on the vacuum wall to
stay in the same position while the camera is moved to each individual tripod to
capture the required images.
17
Figure 4.6: Photo of Vacuum Wall
4.3 Reference Measurement System
A third measurement system used in this analysis is a reference system. This system
was designed and built to handle large pieces of screen material and resembles a large
laser scanner. This system can only accept film that is 1.2m long, which is about
one-third less in size than the vacuum wall described above. In this system, a large
platform scans in X and Y underneath a green laser. The laser is incident on the
material at approximately 5◦ . A detector is placed at the specular reflection angle
(−5◦ ). This captures the reflectivity of the sample at near normal. Over the course
of approximately eight hours, the platform scans the entire sample at 1cm resolution.
The result of this is an XY map of reflectivity over the entire sample.
18
CHAPTER 5
Data Processing Algorithm
5.1 Processing Method for Image Based Method
With all of the data sets collected, the next step is to pull the relevant information
from each set and return a uniformity map of the screen material. There are four
main steps to this process. 1. Crop each image to the region of interest, 2. Calculate
and remove the projector illumination, 3. Determine the angle from the specular
for each pixel in the images, 4. Generate a gain map, 5. Remove the gain map from
the data.
The luminance map of the screen sample is referred to as the raw data. This
also represents B(x, y) in equation 3.3.
The first step is to determine the cropping region for the images. The sample of
screen material is the limiting element in all of the captured images, so the edges
of the screen sample will dictate the cropping region and any necessary rotations.
Since the camera remained stationary for each image, the region defined by the
screen sample will be used for all images.
Even though the projector and camera are fixed for a particular set of images,
each point on the screen has a unique angle of incidence and viewing angle. Using
the geometry of the projector and camera placement, we can calculate the view angle
relative to the specular angle of reflection for each pixel in the images. First a basic
coordinate system is setup such that the screen center is at location (0,0,0). The
camera and the projector point in the +z direction. This allows for each component
to be assigned a coordinate in space and directional vectors to be defined. The
normal vector of the screen is also required, which in this case is (0,0,+1) for every
19
Figure 5.1: Geometry of the Projected and Viewer Rays
location on the screen. Refer to figure 5.1. Using the screen normal vector (dashed
line) and the projected ray, a vector that corresponds to the specular reflection is
calculated for every pixel in the image. The angle between the projected ray and the
specular ray is denoted θi .
These vectors are then used with the camera location
to calculate the angle the viewer ray makes relative to the specular reflection angle,
θs . This is repeated for every location on the screen (in essence every pixel in the
image) to create a map of angles relative to the specular reflection. The result of
this calculation is shown in Figure 5.2.
The next step is to determine the projector illumination profile. In terms of
equation 3.3, this processing step determines the value of P (x, y) based on a combi-
20
Angle from Specular
Angle from Specular
11
32.5
10
32
9
31.5
8
31
7
30.5
6
30
5
29.5
4
3
29
2
28.5
1
28
(a) On Axis
(b) 30◦ Off Axis
Figure 5.2: Map of Angles from the Specular Reflection for each pixel in the image.
nation of two calculations. The first is to remove the gain profile of the white paper
used in the reference image. The gain curve of the white paper, as measured on
the goinometer, is used as a lookup table. For each pixel in the map of angles from
the specular direction, a relative illumination value is selected from the gain curve.
This generates a map of the paper’s contribution to the luminance in the reference
image. This map is divided out from the white reference sample image. We are
now left with the projector illumination profile. As shown in Figure 5.3a, the image
of the white reference sample has gaps between the individual pieces of paper and
must be filtered out. A wide gaussian filter (relative to the size of the paper gaps)
is passed over the image. The filter has a width of 50pixels and a FWHM of 120px.
This filter removes the spikes in the data from the paper gaps but maintains the
profile of the projector illumination. Figure 5.3b shows the reference image and the
projector illumination profile. The illumination profile is then divided out of the
raw screen data pixel by pixel. The result of this division is shown in Figure 5.5b.
The next step is to remove the gain profile of the material from the data. This
gain profile represents G(θi , φi , θv , φv ) in equation 3.3. Again, using the map of
21
White Reference Image (fL)
40
Projector Profile (fL)
40
39
39
38
38
37
37
36
36
35
35
34
34
33
33
32
32
31
31
30
30
(a) White Reference Image
(b) Projector Profile
Figure 5.3
angles generated in a previous step, we interpolate the gain curve measured by the
goniometer for each point in the map of angles. This generates a gain map (shown in
Figure 5.4). This also represents the situation where the screen sample is perfectly
uniform.
This map is now divided from the screen sample data, thus resulting in
U (x, y). The data at this point is also normalized to the mean to provide a percent
uniformity. The final result of the processing is a uniformity map as shown in Figure
5.5c. The red areas indicate sections of the screen that are higher than the average
brightness. The blue areas represent areas of the screen that are lower than the
average brightness. In a perfectly uniform situation, this map would be a single
color or a single shade of gray.
5.2 Processing Method For Reference Systems
The data for the reference system is far easier to process. The data is stored in
a single file that contains the X and Y values for each location in mm. For each
combination of X and Y, a reflectivity value of the material recorded. This data
22
Gain Profile
1.45
1.445
1.44
1.435
1.43
1.425
1.42
1.415
1.41
1.405
Figure 5.4: Gain Map of Screen Sample
Raw Data (fL)
Uniformity Map (%)
Illumination Removed
50
6
1.44
49
1.42
4
48
1.4
47
46
2
1.38
0
45
1.36
44
-2
1.34
43
42
-4
1.32
41
1.3
(a) Raw Data
-6
(b) Removing the projector (c) Uniformity
illumination profile
Screen Sample
Map
of
Figure 5.5: Data Processing Sequence
is the direct uniformity of the material and thus no additional factors need to be
removed from the data. An average reflectivity is calculated and then the data is
normalized to this average, thus resulting in the percent uniformity. This is shown
in Figure 5.6a. As a comparison, the same material has been measured with the
image-based approach. This data is shown in Figure 5.6b. These two figures are
scaled the same and indicate a nearly identical match of each other. The repeating
23
(a) Data from Reference Systems
(b) Data from Image Based Approach
Figure 5.6: Comparison to Reference System
dark blue features on the right side that show up in figure 5.6a also show up on
the right side of figure 5.6b. The ride vertical band near the left side is present in
both figures, but in figure 5.6b, there is more detail from the higher resolution data.
This provides validation that the image based approach is capable of assessing the
uniformity of a screen sample.
24
CHAPTER 6
Results
With all of the image processing completed on the raw data, the information can
be used in a variety of ways to provide guidance about the overall shape of the gain
changes, the best regions for uniformity, any defects and their locations.
The image of uniformity is very useful in itself. A quick visual inspection can
highlight areas of concern. In Figure 6.1, a localized and repeating defect is seen
in the right region of the image. This particular defect is dark and looks like a
comet around the X=650mm location. This defect also repeats at the Y=300mm,
900mm, and 1500mm. Therefore it can be concluded that this defect is inherent in
the casting tool. This results in an immediate rejection of the material and the tool.
However, sometimes the average uniformity of the image is needed. In this
instance, the average horizontal profile and the average vertical profile of the sample
are determined, as shown in Figure 6.2. These plots are quite useful to be able to
match the subtleties of the profiles to the image itself. This can help to confirm
that features in the profiles do match the features in the image. In the case of this
data set, a dark horizontal line on the left side of the screen panel is shown. Also
seen are some alternating bright and dark sections at about X=500mm position.
These down web features may show up as a checkerboard pattern if this material
were to be made into a full sized screen. Further inspection of the horizontal profile
also indicates that the region between X=200mm and X=550mm may be suitable
for a final screen. The profile shows approximately the same brightness at these
locations, therefore there would no perceivable gain step at the seam.
In another example, Figure 6.3 shows some bright and dark vertical bands. These
bands are quantified in the horizontal cross section However, the other aspect to note
25
Uniformity Map - Defects
0
4
200
3
400
Screen Coordinate (mm)
2
600
1
800
0
1000
-1
1200
-2
1400
-3
1600
-4
1800
0
100
200
300
400
500
600
700
800
Screen Coordinate (mm)
Figure 6.1: Example of Localized Defect
on this sample is the down web uniformity, or vertical cross section as shown. This
material would show no signs of a checkerboard pattern.
These images are used as the initial evaluation of the material. If the single panel
data looks promising, the data is used to mimic the screen construction method as
described in the introduction. Some features in the single panel data may not
become apparent until they are repeated with multiple panels seamed together. For
this reason, we can take the single panel data and stitch it together to form a virtual
screen. Using the data similar to Figure 6.2, we can make the virtual screen shown
in Figure 6.4.
With the data, a region of the single panel is chosen to build a final screen.
This provides a method to examine repeating patterns on the screen along with the
potential gain step when two panels are seamed together. With a quick change of
the usable area, we can see what the impact to the final is determined. It would
normally take about 1-2 weeks of trial and error in screen assembly to gather this
data. The uniformity data used in Figure 6.4 was chosen to highlight that gain
step that can occur at a seam. Now that very good data on the screen uniformity is
26
Uniformity Map (+/-10%)
Average Vertical Cross Section
1800
1600
Screen Coordinate (mm)
1400
1200
1000
800
600
400
200
-10
0
10
Percentage Difference around the Average
Average Horizontal Cross Section
10
8
Percentage Difference around the Average
6
4
2
0
-2
-4
-6
-8
-10
100
200
300
400
500
600
700
Screen Coordinate (mm)
Figure 6.2: Average Horizontal and Vertical Profile
27
Uniformity Map (+/-10%)
Average Vertical Cross Section
350
Screen Coordinate (pixles)
300
250
200
150
100
50
-10
0
10
Percentage Difference around the Average
Average Horizontal Cross Section
10
8
Percentage Difference around the Average
6
4
2
0
-2
-4
-6
-8
-10
20
40
60
80
100
120
140
Screen Coordinate (pixles)
Figure 6.3: Example of Good Down Web Uniformity
28
Virtual Screen: 21.5ft x 9.5ft (Scaled to ±10% uniformity)
Figure 6.4: Example of a Virtual Screen
available, this quick simulation process is a main contributor to the final instructions
for how to build a screen. Taking this a step further and projecting this simulated
pattern at real size will provide an even better idea of what features the eye can
pick up on the screen.
Having these uniformity maps allows for a variety of analysis options. While only
a couple of options are presented here, new ways to use the data is being discovered.
Another possible option is to combine on axis and off-axis uniformity maps to fully
simulate how a very large cinema screen might appear to the viewer. In this case,
the screen often fills the viewer’s field of view, such as at a premium large format
theater, and a single on-axis analysis does not provide all the necessary information
to adequately simulate the screen performance.
29
CHAPTER 7
Summary
This method of determining the uniformity of a small section of a cinema screen
allows for a data driven and quantifiable evaluation process that provides the necessary instructions for the construction of a uniform screen. This method is tied
directly to the required specifications of the final product in the fact that it can
prove if a new sample of material meets the requirements or can even drive the
requirements by quantifying what is pleasing to the eye. It is capable of this while
even reducing the cycle time required to gather the data. The data collected with
this method drives the decisions of what a good piece of screen material is and what
region of the replication tool should be used in the final screen assembly.
30
REFERENCES
Coleman, D. and G. Sharp (2013). 54.2: Invited Paper: High Efficiency Polarization Preserving Cinema Projection Screens. SID Symposium Digest of
Technical Papers, 44(1), pp. 748–751. ISSN 2168-0159. doi:10.1002/j.21680159.2013.tb06322.x.
Donohue, J. and N. Fernandes (2004). Closed-loop control maintains arc-lamp stability. Laser Focus World, 40(4), p. 121.
Stover, J. C. (2012). Optical scattering: measurement and analysis. SPIE Press,
Bellingham, Wash. ISBN 9780819492517; 0819492515.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement