| We present several feature-driven illustrative visualization and graphics techniques to enhance the representation of the features of interest in volume datasets. While the magnitude and resolution of real-life datasets keep increasing dramatically, there is a limit on the screen pixel density the human eye can resolve, and a bound on the information a human brain can visually process at any given time. Therefore, we devise techniques to facilitate the perception of the visual information.;First, we propose a GPU-based focus+context framework that uses various standard and advanced magnification lens rendering techniques to magnify the features of interest, while compressing the remaining volume regions without clipping them away completely. Our technique allows the user to interactively manage the available screen area, dedicating more area to the more resolution-important features. A generalization of this concept is multiperspective rendering, which is also studied in our framework to show the spatial relationships of features.;Second, when features are simply magnified, there will always be a limit on the available level of detail and the resolution of the data. To address these shortcomings, we present a technique to extend regular zooms to semantic zooms. Our technique generates the missing detail from any available and plausible high-resolution datasets, using constrained texture synthesis. We demonstrate our approach by ways of a medical application -- the visualization of a human liver -- but its principles readily apply to any scenario, as long as data at all resolutions are available.;The third topic is related to surface texture mapping and synthesis, where we present two methods that preserve both scale and angle. By using global conformal parameterization, the 3D surface texture synthesis problem can be converted to a 2D image synthesis problem. Our multi-scale synthesis method maintains a more uniform area scaling factor. By employing a conformal factor-driven mass-spring relaxation on global parameterization, our second method helps preserve orthogonality and size in texture mapping.;This thesis also seeks to break new grounds in embedding concepts from human perception, cognition, and visual processing into visualization design. We present a rule based color design system to provide better control for task-driven or feature-driven visualization tasks. Our system not only assists in the selection of proper colors, it also helps to avoid poor color mixing in semi-transparent rendering and the apparent change in brightness in color harmonization. Then, inspired by our work on volume rendering and color design, we propose a general multi-layer multi-volume rendering framework. Finally, we investigate the influence and settings of various volume rendering parameters by conducting a user study with 750 participants, assessing the results via conjoint analysis, a promising paradigm to conduct user studies in visualization developed by close collaborators. |
