R&D White Paper
WHP 092
September 2004
Tests of visual acuity to determine the resolution
required of a television transmission system
J.O. Drewery and R.A. Salmon
Research & Development
BRITISH BROADCASTING CORPORATION
BBC Research & Development
White Paper WHP 092
Tests of visual acuity to determine the resolution
required of a television transmission system
John Drewery and Richard Salmon
Abstract
Experiments have been carried out to determine the human eye’s visual acuity,
using fragments of pictures containing textures of various scales, and displayed on
a cathode ray tube. Care has been taken to ensure that the display properties have
not influenced the result.
The results confirm the widely held value of 1 minute of arc, and can be used to
determine the TV standards needed in the coming large-screen display age.
It is concluded that the 1280x720 transmission format will suffice for the
foreseeable future for domestic requirements in Europe.
Additional key words: HDTV flat-screen panel
© BBC 2004. All rights reserved.
White Papers are distributed freely on request.
Authorisation of the Chief Scientist is required for
publication.
© BBC 2004. All rights reserved. Except as provided below, no part of this document may be
reproduced in any material form (including photocopying or storing it in any medium by electronic
means) without the prior written permission of BBC Research & Development except in accordance
with the provisions of the (UK) Copyright, Designs and Patents Act 1988.
The BBC grants permission to individuals and organisations to make copies of the entire document
(including this copyright notice) for their own internal use. No copies of this document may be
published, distributed or made available to third parties whether by paper, electronic or other means
without the BBC's prior written permission. Where necessary, third parties should be directed to the
relevant page on BBC's website at http://www.bbc.co.uk/rd/pubs/whp for a copy of this document.
BBC Research & Development
White Paper WHP 092
Tests of visual acuity to determine the resolution
required of a television transmission system
John Drewery and Richard Salmon
1
Introduction
An earlier (unpublished) set of tests attempted to establish the acuity of the human eye based on the
perception of edges. These were carried out with a television display, using television signals
representing edges varying in step size, mean brightness and hue. The results indicated that a rise
angle of around 1.5 to 3 minutes of arc was just perceptible. From this it is possible to deduce the
number of pixels required for a given size of display at a given distance.
Although the results obtained were valuable, they could be criticised as not relating to real pictures.
Moreover, the methodology did not involve a comparison with a reference, the subject being
required to indicate when they just perceived the loss in edge resolution as the filtering was
increased. So a further series of tests has been carried out involving actual picture material and
direct comparison with a reference, unfiltered picture.
2
Theoretical considerations
In these tests, two types of filter frequency characteristic were used. One was a practical
implementation of a well-known sharp-cut filter (the ITU Rec. 601[1] channel-filter), and the other a
softer filter that avoids ringing or overshoots (the Modified Raised Cosine filter). The horizontal
component of the frequency characteristics of the two filters is given here, in Figures 1 – 4, in both
linear and logarithmic forms. The plots are over the range 0 – 702 cycles/active picture width,
corresponding to 0 – 13.5 MHz, the range over which the Rec. 601 pre-filter is defined (and whilst
the range is arbitrary from the point of view of this experiment, it is a convenient point of reference).
Each of these two filters consists of 43 taps, and may be called a prototype. The Rec. 601 prototype
filter has a rise-time (the 10 – 90 % edge rise-time) of 2 pixels and the Modified Raised Cosine
prototype filter has a rise-time of 4 pixels. The prototype filters were then scaled by interpolation to
yield other filters having arbitrary (integer) rise-times. They were turned into two-dimensional
rectangular filters, having the same rise-time in horizontal and vertical directions, given that the
pixels are square (the tests being conducted on a 1280x720 widescreen raster). If the justperceptible rise-time is converted to a just-perceptible rise-angle a result is obtained that can then be
used for any combination of picture size and viewing distance.
In a preliminary test it was found that the first few steps of filtering could be perceived, not as a
resolution loss, but as a reduction of mean brightness. This could be explained by the fact that the
filtering was being carried out on gamma-corrected signals rather than on signals proportional to
light. So the filtering process was modified to include a conversion to light-proportional signals by
raising the unfiltered signals to a power, filtering and then raising the result to the reciprocal power.
The power chosen was 2.35, being the gamma value now considered as representative[2] for a
cathode ray tube transfer characteristic, and consistent with the way in which television signals are
defined (for example, in ITU Rec. 709[3]).
1
Figures 1-4. Horizontal frequency characteristics of prototype filters
3
Design of the tests
The test material consisted of images like that shown in Figure 5. These comprised two patches of
monochrome picture material, the right being unfiltered and the left filtered. Eleven pictures were
used, chosen to represent different kinds of texture and feature size. These pictures can be seen in
Appendix A. The space underneath contained patches of luminance near black level which acted as a test
like that in a PLUGE generator, enabling the monitor to be set up to match the required test conditions.
Fig. 5. Example of test image
2
Each test image had the left patch filtered by various amounts, thereby creating a sequence for each
filter type and each picture under consideration. The 11 pairs of picture patches were filtered with
both kinds of filter, with 14 (approximately) logarithmically ascending rise-times in the range 2 –
33 pixels.
4
Physical Details of the Test
A “17 inch” HDTV monitor, type JVC, model DT-V1700GC, was used, fed with analogue RGB
signals at 1280x720p/60 from a ATI Radeon 9000 video card running in a PC, operating under the
Linux operating system. The frequency characteristic of this arrangement was judged to be
satisfactory, in that its effect would be negligible in comparison with the test material. The
brightness and contrast were set up according to the specifications in ITU Recommendation 500 [4],
as were the ambient lighting and surrounding conditions.
The monitor and PC were placed 6 metres from the subject, this being the standard distance used in
ophthalmic acuity tests. At this distance, the angle subtended by a pixel on the 720p standard, given
the under-scanned display width of 258 mm, was about 34 mradians or 0.115 minutes of arc and
therefore well below the eye’s acuity limit. This ensured that any artefacts caused by the structure
of the display or of the digital sampling were negligible. As each picture patch was 512 ¥ 432
pixels, the corresponding angular subtense was 17.2 ¥ 14.5 mradians or 0.99 ¥ 0.83 degrees, thus
corresponding to the foveola region of the retina[5].
On the other hand, the PC’s keyboard and VDU were located next to the subject for operational
reasons.
5
Test Methodology
18 observers were used and each was given a quick test of visual acuity using a Snellen chart[6] to
ensure that their visual acuity was not seriously deficient. All the tests were done with glasses if
normally worn. Roughly half of the observers were experienced but, following the results, one was
excluded as giving results well outside the limits of the others.
Unlike the test format specified in ITU Recommendation 500, the (single) subject could control the
presentation of the images from the nearby keyboard, and the VDU displayed instructions on how
to drive the test. After an introductory explanation, the subject proceeded to the first patch pair and
was instructed to move forwards and backwards along the sequence using the cursor keys until the
point was reached where they could just perceive the loss in resolution of the left patch in going
from one step to the next. This step was recorded and then the next patch pair was presented and so
on until all patch pairs had been explored. The first five results were for “training”, and were
ignored and repeated later.
The total time taken for the subject to judge all the patch pairs varied, as they were allowed to take
as much time as needed, but it was typically 20 - 25 minutes.
3
6
Results
Figures 6 and 7 show the just-perceptible rise-angle as a function of picture, for the two filters. The
vertical scale, “Minutes of Arc”, is the angle subtended at the observer by the 10% to 90% points on
a filtered edge when the effects of the filtering on the picture are just perceptible.
Minutes of Arc
1.50
1.25
1.00
0.75
0.50
0.25
s
is
Ra
at
ag
e
W
er
on
Av
M
KW
Co
ll
fa
er
at
W
st
Ra
tH
In
s
er
ch
s
c
au
2
cy
Lu
ur
bo
H
el
Ki
KW
Ca
ar
an
-M
la
el
st
Fu
si
tio
-T
re
tu
Ro
ce
en
or
Fl
on
V
s
of
e
ag
tt
Co
ns
0.00
Fig. 6. Modified Raised Cosine Filter - (Error bars indicate 95% confidence intervals,
i.e. there is a 95% confidence that the true mean lies within the range of the error bars)
Minutes of Arc
1.50
1.25
1.00
0.75
0.50
0.25
1
c.
er
Re
at
e
ag
er
Av
W
60
fa
ll
ch
s
au
at
W
st
on
M
4
tH
In
s
er
H
el
Ki
Fig. 7. Rec. 601 Filter
Ra
c
2
cy
Lu
ur
bo
ar
an
-M
KW
la
el
st
KW
Ca
si
tio
-T
re
tu
Fu
Ro
ce
en
or
Fl
on
V
s
of
e
ag
tt
Co
ns
0.00
It can be seen that the mean just-perceptible rise-angle varies over the set of pictures by a factor of
about 1.4 with both filters, the mean values over all the pictures being shown at the right. The
values for Rec. 601 filtering are somewhat higher than those for raised cosine filtering. This could
be explained by the fact that the ‘ringing’ on edges caused by the Rec. 601 filtering serves to give a
subjective impression of increased sharpness which allows the rise-time to be increased to give the
same degree of apparent sharpness, similar to the Mach bands generated in the eye[7,8].
However, there is not complete correlation between the results for the two filters. For example, for
the picture ‘Future-TV’, the just-perceptible rise-angle of the Rec. 601 filtered image is 1.6 times
that of the raised cosine filtered image, whereas, for the picture ‘Waterfall’, the same ratio is only 1.05.
It is worth noting that these results give just-perceptible rise-angles that are less than half those
obtained from the edge tests. The latter were in the range 1.5 – 3 minutes of arc, as noted above,
whereas here they are in the range 0.7 – 1.2.
The most important result is that, when using the sharp-cut “Rec. 601” filter, the average visible
limit of edge rise-time lies in the range 0.997 to 1.111 minutes of arc, with a confidence of 95%.
Interestingly, the generally accepted figure for visual acuity is approximately 1 minute of arc.
7
Comments on test method
Compared with the previous series of tests, these tests proved to be easier for the subjects, because
of the presence of the unfiltered reference picture on the right hand side. However, it may be that
the choice of step size used in the sequences was inappropriate because the just-noticeable
difference for resolution is not known.
During the conduct of the tests, some observers commented that they found them difficult because
the picture patches were unduly small. However, as previously noted they were of the size
appropriate to the retina’s foveola.
It has been suggested that LCD displays should have been used, representing the display of the
future, rather than the CRT. However, this test is not display specific, since the display is at a
sufficiently great distance that its structure should have no influence on the outcome of the tests.
8
Implications on the TV standard required for different screen sizes
The test results above represents a situation where the transmission channel is the limiting factor,
with the source and display both having higher resolution. Of the two filters tested, the sharp-cut
filter is the most efficient means of filling a transmission channel.
Calculating the average filter rise-time over the 11 pictures for each of the 17 observers, and
knowing that the 10%-90% rise-time of a sharp-cut filter is 44.6% of the reciprocal of its cut
frequency, it is possible to deduce the transmission standard which the average observer, or each
observer separately, requires when viewing displays of different sizes at a given viewing distance.
The 2.7m viewing distance used to generate the charts presented below is the median (and as it
happens, also the mean) viewing distance from a study of 102 domestic arrangements, reported in
June 2004 in another BBC R&D White Paper[9].
Starting from the average result, 1.054 minutes of arc, (and its 95% confidence limits) the
horizontal resolution that would be required at different screen sizes (for widescreen TVs, without
overscan) at this 2.7m median viewing distance has been calculated. The calculation is based
purely on the horizontal resolution supported by the standard, and says nothing about interlace vs
progressive, or other vertical or temporal factors. This is presented in Figure 8, where the coloured
areas indicate the TV standard (based on horizontal sampling structure of the signal) appropriate for
the required resolution. It shows that standard definition 576-line TV is ideal for screens up to 27.5
5
inch diagonal (or 26 inches with allowance for 5% overscan)*, and that the 1280x720 standard is
suitable for screens up to 50 inch diagonal (or 47.5 inches, allowing for overscan).
Fig. 8. Horizontal sampling of TV standard required for different screen sizes, viewed at 2.7m
Figure 9 shows the proportion of individual observers that require a given standard.
100%
Proportion of test subjects requiring TV standard
90%
80%
70%
60%
702x576
1280x720
1440x1080
1920x1080
4k
8k
50%
40%
30%
20%
10%
0%
22
24
26
28
30
32
35
37
40
Screen Size, inches
42
45
50
55
60
Fig. 9. TV Standard required to provide adequate horizontal resolution for 2.7m viewing distance
________________________________________________________________________________________________
*
In a study of CRT overscan undertaken at BBC R&D in 1999, widescreen CRTs were found to have on average 7.8%
overscan on a 702-sample line. 5% might be assumed a reasonable value as we move towards the age of fixed
geometry (non CRT) screens.
6
Figure 9 shows that a significant proportion of viewers are likely to begin to notice a fall-off in
performance in a standard definition, widescreen, transmission at screen sizes of 26-28 inch
diagonal, when viewed at average viewing distances (green in the chart). With the increasing sales
of larger flat-screen TVs which are forecast[10] for the next few years, a significant proportion of the
BBC’s audience will be using larger screens. Under these circumstances it seems unlikely, in the
foreseeable future, that a significant number of viewers will buy screens larger than 45-50 inches,
due to domestic room constraints. Figures 8 and 9 thus indicate that the majority of this new largescreen audience will find 1280 horizontal pixels sufficient as a transmission standard.
It should be noted that this work does not define the resolution required in the display. Because
displays do not have a sharp cut optical transfer function, due to the physical size and shape of the
pixels, an over-sampled display is required to get the best from the broadcast signals. Thus a 720line television image viewed on a 1080-line display will give a better picture than if viewed on a
native 720-line or 768-line display. When considering that RGB triplets are used rather than a
single display cell, it becomes all the more apparent that an over-sampled display is beneficial.
9
Conclusions
The tests described indicate that human vision can distinguish sharpness of optimally filtered TV
signals up to an edge transition width of about 1 minute of arc, as subtended at the eye.
From these results it can be inferred that, for TV sets in the range 26 to 48 inch diagonal, the
1280x720p TV standard would be optimum. If it is reasonable to expect that mean domestic
viewing distances will neither increase or decrease, since the size of rooms in domestic dwellings is
relatively inflexible, it is only the size of screens that needs be considered. There is a general
consensus of opinion in the display industry that the preferred size for large domestic TV screens is
likely to settle at between 37 and 42 inches as the flat-screen revolution matures. It would therefore
appear profligate to propose a TV standard that assumes a significant domestic penetration of
screens over 50 inches diagonal (which in any case might tend to be installed in larger rooms with
consequently larger viewing distances).
However, as indicated above, a pixel count higher than 1280 would be required in the display to
take full advantage of the information content in the transmitted signal.
7
10 References:
1.
ITU Recommendation, 1995. Studio encoding parameters of digital television for
standard 4:3 and wide-screen 16:9 aspect ratios. ITU-R BT.601-5
2.
ROBERTS, A., 1991. Methods of measuring and calculating display transfer
characteristics (gamma). BBC Research Department Report No. RD 1991/6
3.
ITU Recommendation, 2002. Parameter values for the HDTV standards for
production and international programme exchange. ITU-R BT.709-5
4.
ITU Recommendation, 2002. Methodology for the subjective assessment of the quality of
television pictures. ITU-R BT.500-11
5.
HUNT, R.W.G., 1991. Measuring Colour, 2nd Ed.; p. 24. Ellis Horwood, Chichester, UK
6.
STROUSE WATT, W., O.D. How Visual Acuity Is Measured. Macular Degeneration
Support website: http://www.mdsupport.org/library/acuity.html - see also:
http://www.mdsupport.org/snellen.html
7.
RATLIFF, F., 1965. Mach Bands: Quantitative Studies on Neural Networks in the Retina.
Holden-Day, San Francisco
8.
FIORENTINI, A., 1972. Handbook of Sensory Physiology, Vol. VII/4: Visual
Psychophysics (Ed. D. Jameson and L.M. Hurvich); Chapter 8, pp 188-201. SpringerVerlag, New York
9.
TANTON, N.E., 2004. Results of a survey on television viewing distance. BBC R&D
Department, White Paper No. 90
10. SALMON, R.A., 2004. The changing world of TV displays - CRTs challenged by flat-panel
displays. BBC R&D Department, White Paper No. 89
8
Appendix A: Test pictures used in the texture resolution tests
Cottage
Florence Roofs
9
Future-TV
KW Castellations
10
KW Mansion
Kiel Harbour
11
Lucy
Monsters Inc 2
12
RatHaus
Watch
13
Waterfall
14