Resumen de Enhanced perception in volume visualization
Due to the nature of scientic data sets, the generation of convenient visualizations may be a difficult task, but crucial to correctly convey the relevant information of the data. When working with complex volume models, such as the anatomical ones, it is important to provide accurate representations, since a misinterpretation can lead to serious mistakes while diagnosing a disease or planning surgery. In these cases, enhancing the perception of the features of interest usually helps to properly understand the data. Throughout years, researchers have focused on different methods to improve the visualization of volume data sets. For instance, the definition of good transfer functions is a key issue in Volume Visualization, since transfer functions determine how materials are classified. Other approaches are based on simulating realistic illumination models to enhance the spatial perception, or using illustrative effects to provide the level of abstraction needed to correctly interpret the data. This thesis contributes with new approaches to enhance the visual and spatial perception in Volume Visualization. Thanks to the new computing capabilities of modern graphics hardware, the proposed algorithms are capable of modifying the illumination model and simulating illustrative motifs in real time. In order to enhance local details, which are useful to better perceive the shape and the surfaces of the volume, our first contribution is an algorithm that employs a common sharpening operator to modify the lighting applied. As a result, the overall contrast of the visualization is enhanced by brightening the salient features and darkening the deeper regions of the volume model. The enhancement of depth perception in Direct Volume Rendering is also covered in the thesis. To do this, we propose two algorithms to simulate ambient occlusion: a screen-space technique based on using depth information to estimate the amount of light occluded, and a view-independent method that uses the density values of the data set to estimate the occlusion. Additionally, depth perception is also enhanced by adding halos around the structures of interest. Maximum Intensity Projection images provide a good understanding of the high intensity features of the data, but lack any contextual information. In order to enhance the depth perception in such a case, we present a novel technique based on changing how intensity is accumulated. Furthermore, the perception of the spatial arrangement of the displayed structures is also enhanced by adding certain colour cues. The last contribution is a new manipulation tool designed for adding contextual information when cutting the volume. Based on traditional illustrative effects, this method allows the user to directly extrude structures from the cross-section of the cut. As a result, the clipped structures are displayed at different heights, preserving the information needed to correctly perceive them.
