[ IBM Research ]
 
Visual Analysis Home
Perceptual Experiments in Color, Shape, and Texture
We are very interested in how human perception relates to visualization, and in the scientific application of principles of human perception to creating accurate and valid visualizations. The purpose of visualization is to create visual representations that effectively communicate the structure of the data without creating visual artifacts. One area of research that we are currently involved in is the development of colormaps that faithfully represent the structure in the data.
Color Naming
The extraction of high-level color descriptors is an increasingly important problem, as these descriptions often provide a link to image content. When combined with image segmentation, color naming can be used to select objects by color, describe the appearance of the image and even generate semantic annotations. 
Color naming
Our work focuses on the development of computational models for color categorization, color naming and the extraction of color composition in complex scenes.
Perceptual Studies of Colormaps
"colormaps" We have conducted experiments to measure how hue, luminance, and saturation scales represent magnitude information by constructing colormaps which trace carefully controlled paths through HSV and L*a*b* color spaces. The figure to the left shows colormaps that were constructed by tracing trajectories along the hue, luminance, and saturation components of the HSV colorspace.
To measure the ability of these colormaps to faithfully represent data values, we determined the just-noticable differences between points in the colormaps. This was done by measuring the amplitude of a just-detectable "Gaussian target" at 20 different values along the range of each color map. The two figures below show the design of the Gaussian target, and demonstrates how reducing the Gaussian amplitude reduces the detectability of the change.
"perception experiment" "perception experiment"
The results of these experiments show that for the task of effectively representing continuous interval data, luminance-varying colormaps are the best choice, saturation-varying colormaps are somewhat effective, and that hue-varying colormaps are not useful at all. 

Building Perceptual Color Maps for Visualizing Interval Data, Alan D. Kalvin, Bernice E. Rogowitz, Adar Pelah, Aron Cohen, SPIE Conference on Human Vision and Electronic Imaging V, Jan 2000.
PRAVDA
An important aspect of developing visualization applications is in the design of the actual pictorial content. In particular, we have been working on creating more efficient methods to design visualizations, by introducing rules to the process under user control and interaction. To date, the effort has focused on perceptual rules, particularly in color. In the spirit of clever acronyms, from which you fill in the blanks, this approach is called PRAVDA for Perceptual Rule-based Architecture forVisualizing Data Accurately. A rule-based advisory tool for designing colormaps has been developed using these principles. 

A Rule-based Tool for Assisting Colormap Selection, L. Bergman, B. Rogowitz and L. Treinish. Proceedings of the IEEE Computer Society Visualization '95 pp. 118-125, October 1995.
IEEE - Copyright 1995 by IEEE. Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit. To copy otherwise, to republish, to post to servers, or to redistribute to lists, requires prior specific permission and/or fee.