Date of Award

Winter 2010

Project Type


Program or Major

Computer Science

Degree Name

Doctor of Philosophy

First Advisor

Colin Ware


Researchers have argued that perceptual issues are important in determining what makes an effective visualization, but generally only provide descriptive guidelines for transforming perceptual theory into practical designs. In order to bridge the gap between theory and practice in a more rigorous way, a computational model of the primary visual cortex is used to explore the perception of data visualizations.

A method is presented for automatically evaluating and optimizing data visualizations for an analytical task using a computational model of human vision. The method relies on a neural network simulation of early perceptual processing in the retina and visual cortex. The neural activity resulting from viewing an information visualization is simulated and evaluated to produce metrics of visualization effectiveness for analytical tasks. Visualization optimization is achieved by applying these effectiveness metrics as the utility function in a hill-climbing algorithm. This method is applied to the evaluation and optimization of two visualization types: 2D flow visualizations and node-link graph visualizations.

The computational perceptual model is applied to various visual representations of flow fields evaluated using the advection task of Laidlaw et al. The predictive power of the model is examined by comparing its performance to that of human subjects on the advection task using four flow visualization types. The results show the same overall pattern for humans and the model. In both cases, the best performance was obtained from visualizations containing aligned visual edges. Flow visualization optimization is done using both streaklet-based and pixel-based visualization parameterizations. An emergent property of the streaklet-based optimization is head-to-tail streaklet alignment, the pixel-based parameterization results in a LIC-like result.

The model is also applied to node-link graph diagram visualizations for a node connectivity task using two-layer node-link diagrams. The model evaluation of node-link graph visualizations correlates with human performance, in terms of both accuracy and response time. Node-link graph visualizations are optimized using the perceptual model. The optimized node-link diagrams exhibit the aesthetic properties associated with good node-link diagram design, such as straight edges, minimal edge crossings, and maximal crossing angles, and yields empirically better performance on the node connectivity task.