Transfer function optimization for volume visualization based on visibility and saliency
Citation:
Shengzhou Luo, 'Transfer function optimization for volume visualization based on visibility and saliency', [thesis], Trinity College (Dublin, Ireland). School of Computer Science & Statistics, 2016Download Item:
Abstract:
Volume data is widely used in scientific and medical research, and volume visualization techniques aim to provide effective and flexible methods for analyzing and exploring complex structures in such data. However, obtaining clear visualization of features of interest in volume data is still a major challenge. One time-consuming and unitintuitive part of the process is the specification of an appropriate transfer function, which determines the classification and visibility of features. In practice, this is typically achieved using a trial-and-error approach: modifications are made to the transfer function and changes in the resulting visualization are carefully observed in order to inform further modifications to the transfer function. This thesis proposes and investigates novel automated optimization techniques for transfer functions, in order to emphasize features of interest. These techniques exploit information content associated with volume data and objective measures based on visual saliency and visibility in volume visualization. We describe a global optimization and two user-driven refinement methods for modulating transfer functions in order to assist the exploration of volume data. This optimization is dependent on the distribution of the scalar values of the volume data set and is designed to reduce general occlusion and improve the clarity of layers of structures in the resulting images. In addition to view-independent information, we propose a novel view-dependent measure called visibility-weighted saliency in order to assist users in choosing suitable viewpoints and designing effective transfer functions to visualize the features of interest in a volume rendered images. This measure is based on a computational measure of perceptual importance of voxels and the visibility of features in volume rendered images.
Subsequently, we present an automated transfer function optimization method based on the visibility-weighted saliency metric. This method takes into account the perceptual importance of voxels and the visibility of features, and automatically adjusts the transfer function to match the target saliency levels specified by the user. In addition, a parallel line search strategy is presented to improve the performance of the optimization algorithm. Finally, we describe a multivariate visualization approach which modulates focus, emphasizing important information, by adjusting saturation and brightness of voxels based on an importance measure derived from temporal and multivariate information.
Author: Luo, Shengzhou
Advisor:
Dingliana, JohnQualification name:
Doctor of Philosophy (Ph.D.)Publisher:
Trinity College (Dublin, Ireland). School of Computer Science & StatisticsNote:
TARA (Trinity’s Access to Research Archive) has a robust takedown policy. Please contact us if you have any concerns: rssadmin@tcd.ieType of material:
thesisAvailability:
Full text availableMetadata
Show full item recordLicences: