Computer graphic displays offer the experimenter a flexible tool
for the presentation of visual stimuli. Nevertheless, there are
still a number of interesting stimuli which pose technical problems.
A particularly demanding class of stimuli consists of slowly
moving patterns;
for the sake of concreteness, the following discussion will focus
on the production of a particular stimulus, a sinusoidal grating
drifting slowly with a velocity of 0.1 pixel per frame.
Such stimuli might be used in the measurement of spatio-temporal
contrast sensitivity, as described by Kelly (19xx); the technique
described in this paper is applicable to virtually any stimulus
involving the drifting gratings which are commonly used to investigate
motion perception.
We have described a method of combining
digital halftoning with dynamic lookup table programming
to produce drifting plaid patterns. We have found this
method to be superior to other approaches both in flexibility
and freedom from artifacts. Nevertheless, there are a few artifacts
which are unique to this technique, and the remainder of this note
is devoted to their analysis.
This section is divided into four parts. In the first part we consider
the spatial noise introduced by halftoning; this is a cursory
treatment of a subject which is covered in more depth elsewhere
(Ulichney, 1987), but we include it for completeness and
to provide a framework for things to come. In the second
section we analyze what happens to the halftoning noise when
the color table is modified to drift the patterns, and show
how the resulting spatiotemporal noise can be decomposed into
components of flicker and drift. In the third section we analyze
artifacts arising from intensity quantization due to finite
DAC resolution. Finally, we introduce a nonlinear monitor artifact
which to our knowledge has not heretofore been considered.
We hope that the lengthy discussion
of artifacts does not cause the reader to disparage the utility of
the technique; our intention is to demonstrate that the artifacts
are easily calculated, so that researchers using the method
can assure themselves that the artifacts in their stimuli
are negligible.
Halftoning in conjunction with dynamic lookup table modification
is a powerful technique for the generation of moving
stimuli for vision research. Using this technique,
contrast and temporal frequency can be varied with a negligible
amount of computation once a single bitmap image has been
produced. Since only two bit planes are needed to display
a single drifting grating, an 8 bit per pixel display can
be used to generate four component plaids, where each
component of the plaid has independently programmable
contrast and temporal frequency (speed).
(The same display might be used to generate 2 component plaids
where the constituent patterns were halftoned to 2 bits per pixel
instead of 1, reducing the halftoning noise for high
contrast patterns.) Artifacts can be minimized
by careful monitor gamma correction
and by working at pixel contrasts where monitor spatial interactions
are linear. This technique makes it possible to produce complex
motion stimuli which are difficult if not impossible to produce
by other means.
We thank Al Ahumada, Lew Hitchner, and Beau Watson for useful comments on the manuscript. Lee Stone was supported by an NRC research associateship.