Text

theses

Line drawings offer a way to depict 3D shape using a ``minimal'' representation. They can abstract away other features like shading, texture, color, and others while still yielding a vivid sense of shape. Attempts to have computers automatically generate line drawings vary based on the ``definitions'' used to determine line placement. However, the resulting drawings are inconsistent in quality. In order to improve these methods, we use perceptual studies to understand when these methods are effective and when they are not. In the first study, we compare the accuracy of 3D shape percepts from line definitions, artist drawings, and shaded images. We find that line drawings can yield percepts almost as accurate as shaded images, but that no line definition is universally superior to others. We also find some evidence that the quality of these line definitions depends on the geometric context of the line. In the second set of studies, we examine the factors that determine the relevant context of two line definitions: apparent ridges and suggestive contours. We found that the effectiveness of an apparent ridge was largely dependent on the maximal curvature along apparent ridge, while suggestive contours are affected by multiple geometric variables. The final chapter studies these geometric variables in greater detail, and uses them to create a novel classification scheme for suggestive contours. Implementing this classification scheme allows for greater control over which suggestive contours are drawn and separate thresholds for each category, allowing users to include the ``best'' suggestive contours. We also show that this classification scheme allows users to combine suggestive contours with other line definitions to generate more effective line drawings.

ETD_6368

Ph.D.

Includes bibliographical references

by Kevin Sanik

doi:10.7282/T32V2J0Q

ETD doctoral

The author owns the copyright to this work.