10.4 / I. Kawahara
10.4: Measurement of Moving Picture Resolution for Displays
Using Scrolled Sine-Bursts
Isao Kawahara * **
* Image Quality Project, Advanced PDP Development Center Corporation (APDC), Japan.
** Panasonic AVC Networks Company, Matsushita Electric Industrial Co., Ltd, Japan
Abstract
Measurement of moving picture resolution for displays has been
established, featuring a set of scrolled sine-bursts with different
frequencies and different signal levels. For higher accuracy,
ingenious ideas including sub-pixel scrolling are introduced,
showing the definite advantages over response-time-based
approaches. Automated system and human visual perception are
also discussed.
1. Introduction
With the spread of digital cameras with high pixel count, and the
penetration of HDTV digital broadcasting in numerous countries
and regions, consumers have also become more interested in the
resolution of their TV sets. Flat Panel Displays (FPDs) appear to
have replaced many of the CRTs in the market, appealing high
image quality equipped with the latest technologies. Also, various
sizes with “1920x1080” resolution, which is the high-end format
for current broadcasting, are available at moderate prices from
manufacturers.
However, some of FPDs do not have sufficient performance
especially in showing moving pictures, even if they are advertised
as TVs suitable for movies, not as PC monitors. Some
manufactures even claim that their response of panel device being
one of the fastest, so that they could reproduce flawless image
quality in fast motion, disregarding the actual blurred motion
images owing to hold-time effect from driving scheme [1]. In
LCD, with the over driving technologies and new LCD modes,
average response has been improved [2]. However, further efforts
are required before we reach sufficient level as Full-HDTV.
LCD manufacturers, as well as researchers often use 'response
time' to indicate panel performance [3,4,5], though their
measurement are not carried out in a uniform criteria. Some uses
'rise-and-fall response' (TrTf), and others prefer 'Gray-to-Gray'
LCD response to TrTf. In most cases, manufacturers rarely
indicate on what condition they made such measurements.
Therefore, comparison of response times in brochures is almost
pointless.
Among them, MPRT [3] is a response time measurement based on
pursuit camera system, simulating eye tracking for moving
pictures. This means MPRT is straightforward to human
perception and has been accepted by the experts so far, despite the
crucial drawbacks as mentioned in the next session. As of the first
quarter of 2008, standardization process on MPRT is expected to
be completed by ICDM (International Committee for Display
Metrology) before long.
2. Limit of 'Response Time' measurement
Although MPRT is straightforward and has been well accepted by
display experts, it has a fundamental disadvantage arising from
'response time' based measurement itself.
2.1 Benefit of Pursuit Camera System
Unlike recent tendency of approaches using high-speed camera,
MPRT system is a simple and direct emulation of human
perception system, and images captured by the system are exactly
the emulation of what human perceives.
In contrast, high-speed camera approaches are just simulation
only, requiring high speed shuttering, also expecting relatively
slow response of any light emission from target sample displays.
Otherwise, simulated results will not produce reasonable output in
the high-speed camera system.
2.2 Benefit of 'Response Time' Measurement
Computer monitors are available in a wide range of screen sizes,
pixel counts, aspect ratios, frame rates, and so on. One largest
merit of using 'response time' is its simplicity in showing the
results using milliseconds ([ms]) as a unit, regardless of the
differences in all other features of the target displays. For
example, response for an SXGA monitor at 70Hz operation can be
measured in [ms] just the same as for a 1920x1080 HDTV
monitor on 60Hz interlaced source.
2.3 Limit of 'Response Time' Based Measurement
Despite the important advantage mentioned in the section 2.2,
response time has some crucial limits. From an engineering point
of view, a waveforms captured by an MPRT system as a response
to the test chart, is called a 'step response', and the waveform itself
should contain all the information regarding moving picture
performance, as long as the display system is considered as linear
system. However, many of FPDs is not a linear in displaying and
perceiving moving pictures.
More importantly, to degenerate a waveform of a 'step response',
which contains important information, into a mere 'response time',
should end up in discarding substantial information on moving
picture performance. This is exactly what MPRT is doing in
estimating so-called BET or EBET, from the step response [4]. In
0
0 .2
0 .4
0 .6
0 .8
1
1 .2
0 10 20 30 40 5 0
E lapsed T im e (arbitrary unit)
L um inance (N orm alized)
90%
10%
EBET
BET
Figure 1: In this figure, a value of BET (or EBET),
given by MPRT, fails to differentiate two different
step responses.
ISSN/008-0966X/08/3901-0121-$1.00 © 2008 SID SID 08 DIGEST • 121
10.4 / I. Kawahara
general, by selecting just a few points form the step response and
making extrapolation or interpolation to obtain EBET, higher
frequency components will be lost and two different step
responses will give one identical BET or EBET as shown in
Figure 1. This may happen even if the whole display system is
considered as linear.
In essence, 'response time' is equivalent to an approximated value
of frequency response at just one frequency point somewhere in
low to middle range. Therefore, 'response time' is neglecting the
important high frequency components contained in the original
step response. It is also impossible to predict the original
frequency response from a degenerated 'response time' as given in
MPRT, for example.
2.4 Human Perception vs. Response Time
It is reported that owing to so-called 'Mach band effect', blurred
width may look differently according to viewing distance from the
screen. This means that viewing distance in a subjective test,
which is required to validate MPRT results being close to human
perception, should be strictly defined. In other expression, width
of blurred edges are likely to be evaluated differently depending
on the viewing distance. All these indicate that evaluation of
displays in this approach is closely connected with human vision,
and it is difficult to separate display factor alone from human
factors.
In short, 'response time' based measurement is not a sophisticated
approach in evaluating moving picture performance of display
itself.
3. Measurement Using Sine-Bursts
To overcome the issues raised in the section 2, the advanced PDP
development center corporation (APDC) have developed an MTF
related concept using a set of sine-bursts to quantify moving
picture performance, with simple naming as 'Moving Picture
Resolution.' The APDC has also established measurement
procedures and a prototype of an automated system [7,8,9].
The concept has following unique features:
- Effective test patterns using sine-burst
- Intuitive unit of TV-lines
- Reasonable resolution step of 50 TV-lines
- Ingenious sub-pixel scrolling by 6.5ppf
- Easy-to-do subjective test, being robust and accurate
3.1 Test Charts
3.1.1 A Set of Sine-bursts
A set of sine-bursts is a main feature of the APDC test chart as
shown in Figure 2. As stated in the section 2.3.1, 'response time'
retains limited information and is nothing more than an
approximation of a frequency response at only one point in
frequency domain. In contrast, the APDC test chart, containing
sine-bursts from low to high frequencies, is more suitable for
probing detailed frequency response as naturally understood.
3.1.2 Multi-level Sine-Bursts
In addition, sine-bursts in the chart contains three different
background levels as well as three different contrasts for each
frequency, effectively simulating gray levels in natural images.
3.2 Resolution in 'TV-lines'
3.2.1 From Milliseconds to TV-lines
Unit of milliseconds ([ms]), used in 'response time' measurement
may be one benefit as described above, however, no other
particular merit seemed be found. In fact, 'milliseconds' has no
direct association with original still picture resolution, which can
be easily found from the brochures. It is also far from intuitive as
a performance factor of display, since we usually do not directly
differentiate 'time factor' in watching display screens.
In contrast, the APDC method uses 'TV-lines', or 'lines' for short,
as a unit to measure 'Moving Picture Resolution.' It is intuitive
and easy to understand, because it is based on the definition of
still picture resolution of TV, which corresponds to the pixel
count with in the same span of screen height as illustrated in
Figure 3. Also, TV-lines becomes very useful as we can usually
use a most common format like 1920x1080 as input for most test
sample displays for TV use. The perfect score in this case is 1080
TV-lines for all the test sample displays, following the definition
illustrated by Figure 3.
1920
1080
1080
H
3.2.2 Resolution Step of 50 TV-lines
As shown in Figure 2, “APDC Test Chart #1” contains a vertical
patterns stacked with four-cycled sine bursts as shown in Figure 4,
with resolution step of 50 TV-lines. Unlike wedges in
conventional test patterns, these discrete scales make judgment
easier in subjective test, as we can differentiate that subtle
differences appeared on the adjacent sine-bursts. Validation of 50
TV-lines can be found in [6].
3.3 Scrolling Velocity
3.3.1 Typical Speed Representing TV Contents
In general, moving picture performance depends on the speed of
motion. For convenience, we surveyed TV contents and found a
Figure 2: APDC Test Chart #1.
Figure 3:
By definition, still picture
resolution in 1920x1080
panel equals to 1080 TVlines.
Figure 4:
Stacked Four-cycled sinebursts,
with constant band
height of H.
122 • SID 08 DIGEST
10.4 / I. Kawahara
typical speed of “five seconds per screen” representing TV
programs. This is close to a walking speed of a full-shot person on
a 16:9 screen, according to the APDC research.
3.3.2 Higher Accuracy Owing to Sub-pixel Scrolling
Scrolling speed defined as above corresponds to “6.5 pixel per
field” (or 6.5 ppf), in 1920x1080 60Hz format. Having a
fractional figure “5” after the decimal point, signal generators
should output different information between two adjacent fields.
This results equivalently in sub-sampling or double-rate sampling
and improves appearance on Patterns having frequency
component close to Nyquist point on matrix display significantly
as shown in Figure 5.
550 900 1080 [ TV lines ]
Normal Pixel Sampling
Sub-Pixel Sampling
550 900 1080 [ TV lines ]
Normal Pixel Sampling
Sub-Pixel Sampling
3.4 Subjective Test
3.4.1 Accurate and Easy-to-do Test
Subjective test by APDC method is easy to perform just by
following the steps below:
1. Scroll the test chart.
2. Read out maximum resolving point for each patterns.
3. Average the nine results.
4. Average the results from all participants.
In reading out the patterns, we need some special, but simple
attention beyond the critical points, where original “four-line
pattern” often turns into “three dark lines” as shown in Figure 6.
This “three-line pattern” is a false response and mainly caused by
hold time effect of the display, such as typical LCDs. For this
reason, patterns must be checked from lower frequency part (A) to
middle (B) and high frequency part (C) to so as not to overlook
false response.
3.4.2 Robustness in Subjective Test
In a subjective test, decision is made by finding a critical point, or
limit resolution point, where level of response vanishes. This may
sound somewhat ambiguous, as visual acuity and other conditions
like lighting environment are not particularly specified. However,
as long as the observer comes closer enough to the screen, while
keeping minimum distance of distinct vision, valid response as
shown in Figure 6 should look always as valid, and false response
as such. This essentially endorses “Free Viewing Distance” and
makes APDC method far robust against various viewing
conditions compared to 'response time' based measurements
including MPRT.
- 0 .4
- 0 .2
0
0 .2
0 .4
0 .6
0 .8
1
1 .2
0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 7 0 0 8 0 0 9 0 0 1 0 0 0
R esolution [T V -lines]
R esponse (arbitrary unit)
Valid Response False Response
A
B
C
4. Development of Automated System
APDC has also developed a prototype system for automated
measurement as shown in Figure 7. For efficiency in handling
the measurement process, “APDC Test Chart #2” shown as Figure
8 is used.
Figure 5: Sub-pixel sampling improves appearance on
patterns close to Nyquist resolution.
Figure 6: Valid “four-line pattern” (left) and false
resolution of “three dark lines” (right).
Figure 8: “APDC Test Chart #2” for automated
measurement.
Figure 7: Automated Measurement System for
Moving Picture Resolution Developed by the APDC.
SID 08 DIGEST • 123
10.4 / I. Kawahara
BBuursrst tD Deetetecctitoionn
LLeevveel lN Noormrmaalilzizeerr Image
Monitoring
Waveform Check
using
Fourier Analysis
Waveform Check
using
Fourier Analysis
Reference Level
Detection
Reference Level
Detection
Amplitude
Processing
Amplitude
Processing
Phase
Processing
Phase
Processing
DDeeccisisioionn
GO / NG
5. Measured Results
By using the approaches as described above, the automated
system captures images as shown in Figure 10 (Similar images are
perceived by human observers).
As shown in the figure, difference of response for details in
moving picture is clearly discriminated, for example, the
improvement in 120Hz FPD is obvious compared to 60Hz FPD
through this measurement..
Both subjective test and automated system showed excellent
correspondence.
Display A
Display B
(60Hz FPD)
300 TV-lines 350 TV-lines 900 TV-lines
(Marginally) (Not Readable)
OK
Display C
(120Hz FPD)
300 TV-lines 350 TV-lines 900 TV-lines
300 TV-lines 350 TV-lines 550 TV-lines 600 TV-lines
450 TV-lines
OK OK
OK
OK OK OK
NG NG NG
(False Resolution) (Not Readable)
(Not Readable)
NG
6. Discussions
6.1 New Findings on CSF vs. Viewing-Distance
The red line in Figure 11 (A) illustrates a phenomenon known as
CFS, a function explaining that human visual system being most
sensitive to the middle frequency components than to lower or
higher frequency components.
However, while investigating the issues on viewing distance, the
authors have discovered the new fact that the CSF function is not
clearly visible through a linear frequency sweep [9]. Figure 11 (B),
is a linear scaled sweep ending at the Nyquist limit in the right end
of the pattern, where black and white lines are displayed
alternatively line by line. If we come close enough to the screen,
then we should distinguish every detail of the sweep pattern as
Figure 11 (B). Nevertheless, we do not recognize any mountain
shaped boundary on the screen, as long as we use linear sweep of
the right scale as mentioned above.
This also vindicates “Free Viewing Distance” discussion in 3.4.2.
Average luminance over a certain area might be a possible cause of
what has been believed as CFS.
Further studies are required to confirm details on this issue.
7. Conclusions
Moving picture Resolution, the measurement using sine-bursts
established by APDC has numerous advantages over previous
approaches, and was verified to be reasonable and effective through
the experiments and discussions. Being intuitive and easy-to-do
using a unit of TV-lines, the APDC method provides accurate and
quantitative measurement by both subjective test and automated
measurement.
8. References
[1] T.Kurita, A.Saito, I.Yuyama, “Consideration on perceived
MTF of hold type display for moving images,” pp823,
Proc.IDW1998T.
[2] Y. Shimodaira, ‘Fundamental phenomena underlying artifacts
induced by image motion and the solutions for decreasing the
artifacts on FPDs’, SID 2003 Digest, pp. 1034–1037
[3] J.Someya, Y.Igarashi “A Review of MPRT Measurement
Method for Evaluating Motion Blur of LCDs” IDW'04
VHF6/LCT7-1
[4] J. Miseli, J. Lee, J. H. Souk, “Advanced Motion Artifact
Analysis Method for Dynamic Contrast. Degradation Caused
by Line Spreading,” SID ’06, 3.1, pp. 2-5 (2006).
[5] M. Klompenhouwer, ‘Comparison of LCD motion blur
reduction methods using temporal impulse response and
MPRT’, SID 2006, Digest, pp. page 1700-1703
[6] “Resolution Measurement Methods for Digital Cameras”,
CIPA DC-003-Translation-2003, Camera & Imaging Products
Association, pp.6-7, pp. 28-32, (2003)
[7] I. Kawahara, “New Findings on Display Performance in
Large-Sized PDP,” SID ’06, 12.3, pp. 151-154 (2006).
[8] I. Kawahara, “New Method for Measuring Moving Picture
Resolution Suitable for Various Types of FPD”,
EuroDisplay2007, S9-4, pp165-168 (2007)
[9] I. Kawahara, et al, “Measurement and Evaluation of Moving
Picture Resolution: From Milliseconds to TV-Lines”, IDW'07,
VHF1-1, (Invited Paper), (2007)
Figure 9: Main Flow for Making Decisions on Moving
Picture Resolution in Automated System.
Figure 10: Captured Images and Measured Results
from Automated System. (Similar images are
perceived by human eyes.)
(A) (B)
Figure 11: No CSF observed through a linear sweep
(Right) in a right scale. (A): Explained CSF, (B):
Linear Sweep.
124 • SID 08 DIGEST
10.4: Measurement of Moving Picture Resolution for Displays
Using Scrolled Sine-Bursts
Isao Kawahara * **
* Image Quality Project, Advanced PDP Development Center Corporation (APDC), Japan.
** Panasonic AVC Networks Company, Matsushita Electric Industrial Co., Ltd, Japan
Abstract
Measurement of moving picture resolution for displays has been
established, featuring a set of scrolled sine-bursts with different
frequencies and different signal levels. For higher accuracy,
ingenious ideas including sub-pixel scrolling are introduced,
showing the definite advantages over response-time-based
approaches. Automated system and human visual perception are
also discussed.
1. Introduction
With the spread of digital cameras with high pixel count, and the
penetration of HDTV digital broadcasting in numerous countries
and regions, consumers have also become more interested in the
resolution of their TV sets. Flat Panel Displays (FPDs) appear to
have replaced many of the CRTs in the market, appealing high
image quality equipped with the latest technologies. Also, various
sizes with “1920x1080” resolution, which is the high-end format
for current broadcasting, are available at moderate prices from
manufacturers.
However, some of FPDs do not have sufficient performance
especially in showing moving pictures, even if they are advertised
as TVs suitable for movies, not as PC monitors. Some
manufactures even claim that their response of panel device being
one of the fastest, so that they could reproduce flawless image
quality in fast motion, disregarding the actual blurred motion
images owing to hold-time effect from driving scheme [1]. In
LCD, with the over driving technologies and new LCD modes,
average response has been improved [2]. However, further efforts
are required before we reach sufficient level as Full-HDTV.
LCD manufacturers, as well as researchers often use 'response
time' to indicate panel performance [3,4,5], though their
measurement are not carried out in a uniform criteria. Some uses
'rise-and-fall response' (TrTf), and others prefer 'Gray-to-Gray'
LCD response to TrTf. In most cases, manufacturers rarely
indicate on what condition they made such measurements.
Therefore, comparison of response times in brochures is almost
pointless.
Among them, MPRT [3] is a response time measurement based on
pursuit camera system, simulating eye tracking for moving
pictures. This means MPRT is straightforward to human
perception and has been accepted by the experts so far, despite the
crucial drawbacks as mentioned in the next session. As of the first
quarter of 2008, standardization process on MPRT is expected to
be completed by ICDM (International Committee for Display
Metrology) before long.
2. Limit of 'Response Time' measurement
Although MPRT is straightforward and has been well accepted by
display experts, it has a fundamental disadvantage arising from
'response time' based measurement itself.
2.1 Benefit of Pursuit Camera System
Unlike recent tendency of approaches using high-speed camera,
MPRT system is a simple and direct emulation of human
perception system, and images captured by the system are exactly
the emulation of what human perceives.
In contrast, high-speed camera approaches are just simulation
only, requiring high speed shuttering, also expecting relatively
slow response of any light emission from target sample displays.
Otherwise, simulated results will not produce reasonable output in
the high-speed camera system.
2.2 Benefit of 'Response Time' Measurement
Computer monitors are available in a wide range of screen sizes,
pixel counts, aspect ratios, frame rates, and so on. One largest
merit of using 'response time' is its simplicity in showing the
results using milliseconds ([ms]) as a unit, regardless of the
differences in all other features of the target displays. For
example, response for an SXGA monitor at 70Hz operation can be
measured in [ms] just the same as for a 1920x1080 HDTV
monitor on 60Hz interlaced source.
2.3 Limit of 'Response Time' Based Measurement
Despite the important advantage mentioned in the section 2.2,
response time has some crucial limits. From an engineering point
of view, a waveforms captured by an MPRT system as a response
to the test chart, is called a 'step response', and the waveform itself
should contain all the information regarding moving picture
performance, as long as the display system is considered as linear
system. However, many of FPDs is not a linear in displaying and
perceiving moving pictures.
More importantly, to degenerate a waveform of a 'step response',
which contains important information, into a mere 'response time',
should end up in discarding substantial information on moving
picture performance. This is exactly what MPRT is doing in
estimating so-called BET or EBET, from the step response [4]. In
0
0 .2
0 .4
0 .6
0 .8
1
1 .2
0 10 20 30 40 5 0
E lapsed T im e (arbitrary unit)
L um inance (N orm alized)
90%
10%
EBET
BET
Figure 1: In this figure, a value of BET (or EBET),
given by MPRT, fails to differentiate two different
step responses.
ISSN/008-0966X/08/3901-0121-$1.00 © 2008 SID SID 08 DIGEST • 121
10.4 / I. Kawahara
general, by selecting just a few points form the step response and
making extrapolation or interpolation to obtain EBET, higher
frequency components will be lost and two different step
responses will give one identical BET or EBET as shown in
Figure 1. This may happen even if the whole display system is
considered as linear.
In essence, 'response time' is equivalent to an approximated value
of frequency response at just one frequency point somewhere in
low to middle range. Therefore, 'response time' is neglecting the
important high frequency components contained in the original
step response. It is also impossible to predict the original
frequency response from a degenerated 'response time' as given in
MPRT, for example.
2.4 Human Perception vs. Response Time
It is reported that owing to so-called 'Mach band effect', blurred
width may look differently according to viewing distance from the
screen. This means that viewing distance in a subjective test,
which is required to validate MPRT results being close to human
perception, should be strictly defined. In other expression, width
of blurred edges are likely to be evaluated differently depending
on the viewing distance. All these indicate that evaluation of
displays in this approach is closely connected with human vision,
and it is difficult to separate display factor alone from human
factors.
In short, 'response time' based measurement is not a sophisticated
approach in evaluating moving picture performance of display
itself.
3. Measurement Using Sine-Bursts
To overcome the issues raised in the section 2, the advanced PDP
development center corporation (APDC) have developed an MTF
related concept using a set of sine-bursts to quantify moving
picture performance, with simple naming as 'Moving Picture
Resolution.' The APDC has also established measurement
procedures and a prototype of an automated system [7,8,9].
The concept has following unique features:
- Effective test patterns using sine-burst
- Intuitive unit of TV-lines
- Reasonable resolution step of 50 TV-lines
- Ingenious sub-pixel scrolling by 6.5ppf
- Easy-to-do subjective test, being robust and accurate
3.1 Test Charts
3.1.1 A Set of Sine-bursts
A set of sine-bursts is a main feature of the APDC test chart as
shown in Figure 2. As stated in the section 2.3.1, 'response time'
retains limited information and is nothing more than an
approximation of a frequency response at only one point in
frequency domain. In contrast, the APDC test chart, containing
sine-bursts from low to high frequencies, is more suitable for
probing detailed frequency response as naturally understood.
3.1.2 Multi-level Sine-Bursts
In addition, sine-bursts in the chart contains three different
background levels as well as three different contrasts for each
frequency, effectively simulating gray levels in natural images.
3.2 Resolution in 'TV-lines'
3.2.1 From Milliseconds to TV-lines
Unit of milliseconds ([ms]), used in 'response time' measurement
may be one benefit as described above, however, no other
particular merit seemed be found. In fact, 'milliseconds' has no
direct association with original still picture resolution, which can
be easily found from the brochures. It is also far from intuitive as
a performance factor of display, since we usually do not directly
differentiate 'time factor' in watching display screens.
In contrast, the APDC method uses 'TV-lines', or 'lines' for short,
as a unit to measure 'Moving Picture Resolution.' It is intuitive
and easy to understand, because it is based on the definition of
still picture resolution of TV, which corresponds to the pixel
count with in the same span of screen height as illustrated in
Figure 3. Also, TV-lines becomes very useful as we can usually
use a most common format like 1920x1080 as input for most test
sample displays for TV use. The perfect score in this case is 1080
TV-lines for all the test sample displays, following the definition
illustrated by Figure 3.
1920
1080
1080
H
3.2.2 Resolution Step of 50 TV-lines
As shown in Figure 2, “APDC Test Chart #1” contains a vertical
patterns stacked with four-cycled sine bursts as shown in Figure 4,
with resolution step of 50 TV-lines. Unlike wedges in
conventional test patterns, these discrete scales make judgment
easier in subjective test, as we can differentiate that subtle
differences appeared on the adjacent sine-bursts. Validation of 50
TV-lines can be found in [6].
3.3 Scrolling Velocity
3.3.1 Typical Speed Representing TV Contents
In general, moving picture performance depends on the speed of
motion. For convenience, we surveyed TV contents and found a
Figure 2: APDC Test Chart #1.
Figure 3:
By definition, still picture
resolution in 1920x1080
panel equals to 1080 TVlines.
Figure 4:
Stacked Four-cycled sinebursts,
with constant band
height of H.
122 • SID 08 DIGEST
10.4 / I. Kawahara
typical speed of “five seconds per screen” representing TV
programs. This is close to a walking speed of a full-shot person on
a 16:9 screen, according to the APDC research.
3.3.2 Higher Accuracy Owing to Sub-pixel Scrolling
Scrolling speed defined as above corresponds to “6.5 pixel per
field” (or 6.5 ppf), in 1920x1080 60Hz format. Having a
fractional figure “5” after the decimal point, signal generators
should output different information between two adjacent fields.
This results equivalently in sub-sampling or double-rate sampling
and improves appearance on Patterns having frequency
component close to Nyquist point on matrix display significantly
as shown in Figure 5.
550 900 1080 [ TV lines ]
Normal Pixel Sampling
Sub-Pixel Sampling
550 900 1080 [ TV lines ]
Normal Pixel Sampling
Sub-Pixel Sampling
3.4 Subjective Test
3.4.1 Accurate and Easy-to-do Test
Subjective test by APDC method is easy to perform just by
following the steps below:
1. Scroll the test chart.
2. Read out maximum resolving point for each patterns.
3. Average the nine results.
4. Average the results from all participants.
In reading out the patterns, we need some special, but simple
attention beyond the critical points, where original “four-line
pattern” often turns into “three dark lines” as shown in Figure 6.
This “three-line pattern” is a false response and mainly caused by
hold time effect of the display, such as typical LCDs. For this
reason, patterns must be checked from lower frequency part (A) to
middle (B) and high frequency part (C) to so as not to overlook
false response.
3.4.2 Robustness in Subjective Test
In a subjective test, decision is made by finding a critical point, or
limit resolution point, where level of response vanishes. This may
sound somewhat ambiguous, as visual acuity and other conditions
like lighting environment are not particularly specified. However,
as long as the observer comes closer enough to the screen, while
keeping minimum distance of distinct vision, valid response as
shown in Figure 6 should look always as valid, and false response
as such. This essentially endorses “Free Viewing Distance” and
makes APDC method far robust against various viewing
conditions compared to 'response time' based measurements
including MPRT.
- 0 .4
- 0 .2
0
0 .2
0 .4
0 .6
0 .8
1
1 .2
0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 7 0 0 8 0 0 9 0 0 1 0 0 0
R esolution [T V -lines]
R esponse (arbitrary unit)
Valid Response False Response
A
B
C
4. Development of Automated System
APDC has also developed a prototype system for automated
measurement as shown in Figure 7. For efficiency in handling
the measurement process, “APDC Test Chart #2” shown as Figure
8 is used.
Figure 5: Sub-pixel sampling improves appearance on
patterns close to Nyquist resolution.
Figure 6: Valid “four-line pattern” (left) and false
resolution of “three dark lines” (right).
Figure 8: “APDC Test Chart #2” for automated
measurement.
Figure 7: Automated Measurement System for
Moving Picture Resolution Developed by the APDC.
SID 08 DIGEST • 123
10.4 / I. Kawahara
BBuursrst tD Deetetecctitoionn
LLeevveel lN Noormrmaalilzizeerr Image
Monitoring
Waveform Check
using
Fourier Analysis
Waveform Check
using
Fourier Analysis
Reference Level
Detection
Reference Level
Detection
Amplitude
Processing
Amplitude
Processing
Phase
Processing
Phase
Processing
DDeeccisisioionn
GO / NG
5. Measured Results
By using the approaches as described above, the automated
system captures images as shown in Figure 10 (Similar images are
perceived by human observers).
As shown in the figure, difference of response for details in
moving picture is clearly discriminated, for example, the
improvement in 120Hz FPD is obvious compared to 60Hz FPD
through this measurement..
Both subjective test and automated system showed excellent
correspondence.
Display A
Display B
(60Hz FPD)
300 TV-lines 350 TV-lines 900 TV-lines
(Marginally) (Not Readable)
OK
Display C
(120Hz FPD)
300 TV-lines 350 TV-lines 900 TV-lines
300 TV-lines 350 TV-lines 550 TV-lines 600 TV-lines
450 TV-lines
OK OK
OK
OK OK OK
NG NG NG
(False Resolution) (Not Readable)
(Not Readable)
NG
6. Discussions
6.1 New Findings on CSF vs. Viewing-Distance
The red line in Figure 11 (A) illustrates a phenomenon known as
CFS, a function explaining that human visual system being most
sensitive to the middle frequency components than to lower or
higher frequency components.
However, while investigating the issues on viewing distance, the
authors have discovered the new fact that the CSF function is not
clearly visible through a linear frequency sweep [9]. Figure 11 (B),
is a linear scaled sweep ending at the Nyquist limit in the right end
of the pattern, where black and white lines are displayed
alternatively line by line. If we come close enough to the screen,
then we should distinguish every detail of the sweep pattern as
Figure 11 (B). Nevertheless, we do not recognize any mountain
shaped boundary on the screen, as long as we use linear sweep of
the right scale as mentioned above.
This also vindicates “Free Viewing Distance” discussion in 3.4.2.
Average luminance over a certain area might be a possible cause of
what has been believed as CFS.
Further studies are required to confirm details on this issue.
7. Conclusions
Moving picture Resolution, the measurement using sine-bursts
established by APDC has numerous advantages over previous
approaches, and was verified to be reasonable and effective through
the experiments and discussions. Being intuitive and easy-to-do
using a unit of TV-lines, the APDC method provides accurate and
quantitative measurement by both subjective test and automated
measurement.
8. References
[1] T.Kurita, A.Saito, I.Yuyama, “Consideration on perceived
MTF of hold type display for moving images,” pp823,
Proc.IDW1998T.
[2] Y. Shimodaira, ‘Fundamental phenomena underlying artifacts
induced by image motion and the solutions for decreasing the
artifacts on FPDs’, SID 2003 Digest, pp. 1034–1037
[3] J.Someya, Y.Igarashi “A Review of MPRT Measurement
Method for Evaluating Motion Blur of LCDs” IDW'04
VHF6/LCT7-1
[4] J. Miseli, J. Lee, J. H. Souk, “Advanced Motion Artifact
Analysis Method for Dynamic Contrast. Degradation Caused
by Line Spreading,” SID ’06, 3.1, pp. 2-5 (2006).
[5] M. Klompenhouwer, ‘Comparison of LCD motion blur
reduction methods using temporal impulse response and
MPRT’, SID 2006, Digest, pp. page 1700-1703
[6] “Resolution Measurement Methods for Digital Cameras”,
CIPA DC-003-Translation-2003, Camera & Imaging Products
Association, pp.6-7, pp. 28-32, (2003)
[7] I. Kawahara, “New Findings on Display Performance in
Large-Sized PDP,” SID ’06, 12.3, pp. 151-154 (2006).
[8] I. Kawahara, “New Method for Measuring Moving Picture
Resolution Suitable for Various Types of FPD”,
EuroDisplay2007, S9-4, pp165-168 (2007)
[9] I. Kawahara, et al, “Measurement and Evaluation of Moving
Picture Resolution: From Milliseconds to TV-Lines”, IDW'07,
VHF1-1, (Invited Paper), (2007)
Figure 9: Main Flow for Making Decisions on Moving
Picture Resolution in Automated System.
Figure 10: Captured Images and Measured Results
from Automated System. (Similar images are
perceived by human eyes.)
(A) (B)
Figure 11: No CSF observed through a linear sweep
(Right) in a right scale. (A): Explained CSF, (B):
Linear Sweep.
124 • SID 08 DIGEST
