Accuracy in scientific visualisation

Quite often, accuracy is a neglected issue in scientific visualization. Indeed, in most of the visualizations there are two wrong assumptions: first, that the data visualized is accurate. Second, that the visualization process is exempt from errors. On these basis, the objectives of this thesis are three-fold:

First, to understand the implications of accuracy in scientific visualization. It is important to analyse the sources of errors during visualization, and to establish mechanisms that enable the characterization of the accuracy. This learning stage is crucial for a sucessful scientific investigation.

Second, to focus on visualization features that, besides enabling the visualization of the data, give users an idea of its accuracy. The challenging aspect in this case is the use of appropriate visual paradigms. In this respect, the awareness of how human beings create and process a mental image of the information visualized is important.

Thrid and most important, the development of more accurate versions of visualization techniques. By understanding the issue of accuracy concerning a particular technique, there is a high probability to reach to a proposal of new improvements.

There are three techniques under study in this thesis: contouring, isosurfacing and particle tracing. All these are widely used in scientific visualization. That is why they have been chosen. For all of them, the issue of showing accuracy to users is discussed. In addition, two new accurate versions of contouring and isosurfacing techniques have been presented. The new contouring method is for data defined over rectangular grids and assumes that the data vary linearly along the edges of the cell. The new isosurfacing method is an improvement of the Marching-Cubes method. Some aspects of this classic approach are clarified, and even corrected.