next up previous
Next: Ideal Boundary Characterization Up: Semi-Automatic Generation ... Previous: Introduction

Related Work

Two methods have been proposed for assisting the user in the exploration of possible transfer functions. He et al. [7] use genetic algorithms to breed a good transfer function for a given dataset. Judging from small thumbnail renderings, the user picks desirable transfer functions from an automatically generated population, until the iterative process of image selection and transfer function inter-combination converges. Alternatively, the system can run automatically by using some user-specified objective function (entropy, energy, or variance) to evaluate rendered images. Marks et al. [13] address the problem of ``parameter tweaking'' in general, with applications including light placement for rendering, motion control for articulated figure animation, as well as transfer functions in direct volume rendering. The goal is to create a visual interface to the complex parameter space by using an image difference metric to arrange renderings from a wide variety of transfer functions into a ``design gallery'', from which the user selects the most appealing rendering. While both of these methods reportedly succeed in finding useful transfer functions, and while they both allow the user to inspect the transfer function behind a rendering, the systems are fundamentally designed for finding good renderings, not for finding good transfer functions. Both processes are entirely driven by analysis of rendered images, and not of the dataset itself. Rather than having an high-level interface to control the transfer function, the user has to choose a transfer function from among those randomly generated, making it hard to gain insight into what makes a transfer function appropriate for a given dataset.

Other visualization tools have been described which are more driven by the data itself. Bergman et al. [3] describe a perceptually informed rule-based method for colormap selection which takes into account the data's spatial frequency characteristics and the purpose of the visualization. Closer to the goal of the current paper is the contour spectrum, described by Bajaj et al. [1], which helps the user find isovalues for effective isosurface volume visualizations of unstructured triangular meshes. By exploiting the mathematical properties of the mesh, important measures of an isosurface such as surface area and mean gradient magnitude can be computed with great efficiency, and the results of these measurements are integrated into the same interface which is used to set the isovalue. By providing a compact visual representation of the metrics evaluated over the range of possible isovalues, the user can readily decide, based on their rendering goals, which isolevel to use. The importance to the current paper is that the contour spectrum is a good example of how an interface can use measured properties of the data to guide the user through the parameter space controlling the rendering.


next up previous
Next: Ideal Boundary Characterization Up: Semi-Automatic Generation ... Previous: Introduction
Gordon Kindlmann
1999-07-25