The Evaluation of Perceptual Effectiveness of Isosurface Rendering-based Uncertainty Visualization Techniques for Volumetric Scalar Data

Abstract

Many techniques have been proposed to convey uncertainty in visualization. However, little research has been reported on the evaluation of their effectiveness.We present a user study that evaluates the perceptual effectiveness of six (four new and two existing) isosurface rendering-based uncertainty visualization techniques. For every technique, we consider its four effectiveness aspects: identification of the data, identification of the uncertainty, visual overload and brightness. There are thirty users participated in the user study and statistical analysis has been made for this study. Our analysis suggested that the two existing uncertainty visualization techniques appear to be the most advantageous in all the evaluated techniques. Both of them have high scores in all the four aspects of the effectiveness. In terms of the new techniques, the transparency technique appears to be the most promising. Whilst the remaining three new techniques may have some utility in certain aspects of the effectiveness, they are less useful in other aspects of the effectiveness. Additionally, a surprising result we have found is that adding auxiliary grid lines as background is not guaranteed to enhance the participants' perception to the errors depicted by the transparency. Conversely, it may lead to visual overload that increases the difficulty to recognize data. We believe that these findings can be useful for future uncertainty visualization design.