Perceptual Visualization Architecture

This project involves a fundamental investigation of the strengths and limitations of the low-level human visual system. We seek to identify image properties that we can see quickly, to test how these properties interact with one another, and to determine how to harness these properties. Understanding how we perceive the world around us will benefit many research areas including information display, image generation, image analysis, and the simulation of vision. One application of this knowledge is an important area of computer graphics: the visualization of large, complex, multidimensional datasets.

Visualization is the conversion of collections of strings and numbers into pictures that a viewer can use to "see" values, relationships, and structure inherent in their datasets. Multidimensional data visualization presents the dual problems of size and dimensionality. The number of sample points within the dataset is very large; moreover, each sample point contains multiple independent readings or attributes. Viewers want to visualize some or all of this information simultaneously in a single display. The goal of this proposal is to design visualizations that support rapid and accurate exploration and analysis. To do this, we must display all of the data without overwhelming the viewer's visual system.

A solution to this problem is the construction of a perceptual visualization architecture. First, a psychophysical experiments will be used to test how the low-level visual system perceives three fundamental visual features during high-speed display: color, texture, and motion. Results from these studies will form a collection of perceptual guidelines that describe how to combine color, texture, and motion patterns to represent information in an underlying dataset. A "visualization assistant" based on artificial intelligence search algorithms will be constructed on top of these guidelines. This assistant will help viewers choose perceptually optimal methods of converting their data into effective visualizations. Data will be displayed in ways that harness the strengths and avoid the limitations of the viewer's visual system. The result is a set of images that allow viewers to perform rapid, accurate, and effective exploration and analysis.

Participants

Christopher G. Healey (PI)
James T. Enns (PI, University of British Columbia)

Dan Huber (MS candidate)
Geniva Liu (PhD candidate, University of British Columbia)

Publications

• Visualizing Market Data, IEEE Internet Computing (2001)
• Building a Perceptual Visualization Architecture, BIT (2000)
• Large Datasets at a Glance: Combining Textures and Colors in Scientific Visualization, IEEE TVCG (1999)
• High-Speed Visual Estimation Using Preattentive Processing, ACM TOCHI (1996)
• Visualizing Real-Time Multivariate Data Using Preattentive Processing, ACM TOMACS (1995)
  
  
• Sensitivity to 3D Orientation in Textured Surfaces, Psychonomics 2000
• Oriented Texture Slivers: A Technique for Local Value Estimation of Multiple Scalar Fields, GI 2000
• Fundamental Issues of Visual Perception for Effective Image Generation, SIGGRAPH 99 Course
• Building Perceptual Textures to Visualize Multidimensional Datasets, Vis '98
• Applications of Visual Perception in Computer Graphics, SIGGRAPH 98 Course
• Volume Rendering of Abdominal Aortic Aneurysms, Vis '97
• On the Use of Perceptual Cues and Data Mining for Effective Visualization of Multidimensional Datasets, GI '98
• Choosing Effective Colours for Data Visualization, Vis '96
• Harnessing Preattentive Processes for Multivariate Data Visualization, GI '93