Perceptual Invariance of Nonlinear Focus+Context Transformations
Keith Lau, Ronald A. Rensink, and Tamara M. Munzner, Department of Computer Science, University of British Columbia, Vancouver BC, Canada.
Proceedings of the 1st Symposium on Applied Perception in Graphics and Visualization, 65-72. 2004.
[SIGGRAPH 2004 / APGV; Los Angeles, CA, USA.]
Focus+Context techniques are commonly used in visualization systems to simultaneously provide both the details and the context of a
particular dataset. This paper proposes a new methodology to empirically investigate the effect of various Focus+Context transformations on human perception. This methodology is based on the
shaker paradigm, which tests performance for a visual task on an
image that is rapidly alternated with a transformed version of itself.
An important aspect of this technique is that it can determine two
different kinds of perceptual cost: (i) the effect on the perception
of a static transformed image, and (ii) the effect of the dynamics
of the transformation itself. This technique has been successfully
applied to determine the extent to which human perception is invariant to scaling and rotation [Rensink 2004]. In this paper, we
extend this approach to examine nonlinear fisheye transformations
of the type typically used in a Focus+Context system. We show that
there exists a no-cost zone where performance is unaffected by an
abrupt, noticeable fisheye transformation, and that its extent can be
determined. The lack of perceptual cost in regards to these sudden
changes contradicts the belief that they are necessarily detrimental
to performance, and suggests that smoothly animated transformations between visual states are not always necessary. We show that
this technique also can map out low-cost zones where transformations result in only a slight degradation of performance. Finally,
we show that rectangular grids have no positive effect on performance, acting only as a form of visual clutter. These results therefore demonstrate that the perceptual costs of nonlinear transformations can be successfully quantified. Interestingly, they show that
some kinds of sudden transformation can be experienced with minimal or no perceptual cost. This contradicts the belief that sudden
changes are necessarily detrimental to performance, and suggests
that smoothly animated transformations between visual states are
not always necessary.
Back to main publications list.