Interactive Editing and Automatic Evaluation of Direct Volume Rendered Images

PhD Thesis Proposal Defence


Title: "Interactive Editing and Automatic Evaluation of Direct Volume Rendered
Images"

by

Mr. Yingcai Wu


Abstract:

The volume datasets from real applications such as medical imaging and 
computational fluid dynamics often contain multiple sophisticated 
structures. Because of the occlusion of 3D objects, revealing all these 
structures and their 3D spatial relations simultaneously is very 
challenging. Direct Volume Rendering is a powerful volume visualization 
method which allows users to visually explore the volume datasets in a 
highly flexible manner. Despite the powerful capability of direct volume 
rendering for exploring volume data, its inherent complexity of specifying 
rendering parameters often results in a tedious and non-intuitive 
visualization process. In addition, because of its complicated ray casting 
and compositing process, its results (i.e., Direct Volume Rendered 
Images) usually contain some misleading information such as artifacts and 
depth ambiguity, which makes the visualization unreliable and ineffective 
for volume exploration.

In this thesis proposal, we present three methods for improving the 
intuitiveness and effectiveness of direct volume rendering as follows.
1). An editing framework for direct volume rendered images, allowing users to 
interactively explore complex volumetric datasets by directly editing 
direct volume rendered images. Users can intuitively fuse multiple 
features in distinct direct volume rendered images, remove any feature 
from a direct volume rendered image, or blend two direct volume rendered 
images.
2). A palette-style volume visualization method, which can 
automatically store and systematically organize intermediate results 
created during a volume visualization process, such that users can locate 
their desired results quickly and generate a new result based on the 
editing framework. Moreover, users can always keep aware of what they have 
explored so far and so exploration redundancy can be significantly 
reduced.
3). A set of quantitative effectiveness measures, i.e., 
distinguishability, edge consistency, contour clarity, and depth coherence 
measures, to evaluate the effectiveness of a direct volume rendered image 
or a whole visualization process from different perspectives. The 
quantified effectiveness can be provided to users at different levels of 
detail, such that users can be informed when misleading or ambiguous 
information is introduced in a visualization process. With these three 
proposed methods, a comprehensive volume visualization system has been 
developed, enabling users to interactively editing, intuitively 
organizing, and effectively evaluating direct volume rendered images.


Date:     		Friday, 29 May 2009

Time:                   4:00pm-6:00pm

Venue:                  Room 3405
 			lifts 17-18

Committee Members:      Dr. Huamin Qu (Supervisor)
 			Dr. Pedro Sander (Chairperson)
 			Dr. Albert Chung
 			Dr. Chi-Keung Tang


**** ALL are Welcome ****