A while back, I posted the following query:
---------------------------------------------------------------
From: "Carlo L. Tiana" <carlo@vision.arc.nasa.gov>
To: cvnet@skivs.ski.org
Subject: Image jitter data?
Cc: carlo@vision.arc.nasa.gov
I am looking for pointers to papers or reports on the amount of image
vibration that might be acceptable to a viewer to perform specific tasks.
I am interested in specific results if they are available or techniques that
one might use to measure this.
My specific application is to determine how much image stabilization is
"enough" to perform certain tasks, that range from viewing the local news
when captured by cameras flown on a helicopter, to reading the licence plate
on OJ's car from such helicopter. Thus, stable "enough" can mean different
things to different viewers - in one case the perceived image quality is
more important, in the latter case there is a specific task to accomplish.
Any help would be very much appreciated.
---------------------------------------------------------------
I received a lot of useful replies, containing references, suggestions and
other ideas, as well as a lot of requests for a summary. Here's what I
hope are all the references I received. Thanks for all the help.
Carlo Tiana
FLIR Systems, Inc.
(503) 684-3731
---------------------------------------------------------------
Levi, Sharma, & Klein (1997). Feature integration in pattern perception.
PNAS, 94, 11742-11746.
American National Standard for Human Factors Engineering of Computer
Terminal Workstations (ANSI/HFS 100-1988; available from the Human
Factors and Ergonomics Society, Santa Monica, CA, 310/394-1811,
hfes@compuserve.com, $25.00).
The standard contains a specification for acceptable jitter limits
for CRT visual displays. The specification was based on spatial and
temporal frequency limits of human visual; however, the standard
does not discuss the research work per se.
McKee & Westheimer 1975 (about 3deg/s for vernier acuity)
Ciuffreda KJ, J Kellndorfer, D Rumpf. Contrast and accommodation. In:
LW Stark, G Obrecht, eds. Presbyopia. Recent Research and Reviews from
the Third International Symposium. New York: Professional Press, 1987.
pp. 116-122.
Flitcroft, DI. Effects of temporal frequency on the contrast
sensitivity of the human accommodation system. Vision Research. 1988,
28(2), 269-278.
Hung GK, Ciuffreda, KJ. Accommodative responses to eccentric and
laterally-oscillating targets. Ophthalmic and Physiological Optics
1992, 12, 361-364.
Badcock, D.R. & Wong, T.L. Resistance to positional noise in human vision.
Nature. 243, 554-555. 1990.
Badcock, D.R. & Wong, T. The sensitivity of separation discrimination to
spatiotemporal jitter. Vision Research. 30 : pp. 1555-1560. 1990.
Badcock, D.R., Wong, T.L. and Coutant, B.E. The impact of spatiotemporal
jitter on binocular separation and depth discrimination. Vision
Research. 31 : pp. 247-252. 1991.
Badcock, D.R. & Sullivan, K. Resistance to positional noise scales with
target size. Vision Research 34, pp. 1327-1330, 1994.
Wells,MJ & Griffin, MJ (1984). Benefits of helmet-mounted display image
stabilization under whole-body vibration. Aviation, Space and
Environmental Medicine, vol 55(1), 13-18.
Wells,MJ & Griffin, MJ (1984). Flight trial of a helmet-mounted display
image stabilization system. Aviation, Space and Environmental
Medicine, vol 58, 19-22.
"Tracking Performance on a Visual Display Apparently Vibrating at One to
Ten Hertz," Huddleston, J.H.F.; J. of Applied Psychology, 1970, v 34
no 5, pp401-405.
"Effect of the Amplitude of Apparent Vibration, Brightness, and Type Size
of Numeral Reading," Crook, M.N. et al, TR AF6246, WPAFB, Ohio, 1950
"Some Effects of Vibration upon Visual Performance," Dennis, J.F., same
journal 1965 vol 49, pp245-252
... [relevant research in the ] late 1970s and the 1980s ...
resulted in the publication of
vibration standards such as ISO 2631. A review of such work and the
resulting guidelines may be found in Griffin, M.J, Handbook of Human
Vibration, 1990, (Academic Press). Several papers are also referenced
in Boff and Lincoln.
---------------------------------------------------------------
And here are at least some of the comments on the query:
If the foveal image is moving too quickly then ocular accommodation
will be poor and will tend to move towards the individual resting
position. Depending on the distance of the target in relation to the
resting focus position, the target will be out of focus. If the amount
of defocus is larger than the depth of focus of the eye then the image
will appear blurred which may lead to poorer performance at the task
(e.g. resolving small letters on a number plate). Some references are:
I am interested to see your enquiry about image jitter on the cvnet.
This is an area on which I have worked for many years. As you may be
aware the effects of vibration on vision were widely researched in the
late 1970s and the 1980s. This work resulted in the publication of
vibration standards such as ISO 2631. A review of such work and the
resulting guidelines may be found in Griffin, M.J, Handbook of Human
Vibration, 1990, (Academic Press). Several papers are also referenced
in Boff and Lincoln.
However the major complication of this research has always been that
experimental work involving one visual task is not necessarily
generalisable to another task ! Therefore a different line of research
has attempted to simulate the legibility of jittering displays. Since
the loss in legibility is caused primarily by the sluggish responses
of the photochemicals on the human retina (giving motion blur) this is
a practical solution to a complex problem.
In the US CSERIAC have a MAC tool by the name of CASHE-PVS (?) that
adopts such an approach (see
http://cseriac.udri.udayton.edu/products). I have also worked for over
10 years on a similar tool (VISTA). Both adopt a similar approach. The
user imports the type of display or scene available to the observer,
together with an indication of the type of vibration (frequency and
amplitude) likely to be encountered. The software then simulates the
likely legibility of the scene during vibration/jitter by blurring the
image. It is left to the user to assess if the task is achievable in
spite of the image degradation. I have always preferred this more
flexible approach.
We have an ongoing research programme to assess the legibility of
aircraft displays during vibration/jitter. Other parts of the company
have also worked on image stabilisation systems which stabilise an
image in electronics on the basis of accelerometer input (i.e.
vibrating aircraft, vibrating camera but stabilised display).
I did this using an early version of the ViDEOS vision model. Its an Army
report, that I can try to find for you if needed. We got some vibration
data and used it to build a filter to blur text. We compared the blurred
to unblurred text in JNDs. It worked pretty well and the predictions were
definitely ball park. I would be very happy to see you try the same thing
using the newer vision model and our interface.
You could easily set up a quick task to verify the predictions. ...
The only issue is getting some vibration data that
is relevant to your task ...
I don't know of any reports but I'd naively guess that, for applied
purposes, there might be a rather simple relation between the energy
content (in the spatial domain) of the image as a function of spatial
frequency (ESF) and jitter (J). The idea being ESF*J=constant.
The tinkerer/scientist in me sais this:, Get a good loudspeaker, cut off the
membrane, attach a light rod to the magnet and connect the other end
to a picture. Feed the loudspeaker (maybe via a power amp) by the signal
from a waveform generator, esp. sinewaves. This way you will be able to
obtain a frequency characteristic (sort of). You can vary the amplitude,
if necessary, by varying the viewing distance. I have seen it done in a
series of experiments on frog muscle spindles. I can get more info if
you need it.
I would not do such experiments using a computer screen. Being tied to
a fixed frame rate is extremely limiting. Maybe an analog X-Y-scope could
work, but I have no experience with it.