The increasing usage of volumetric imaging modalities, such as magnetic resonance imaging (MRI) and computed tomography (CT) in areas such as medicine and nondestructive evaluation (NDE) has placed a great importance on 3D visualization techniques. This growth of volume data in the form of cross sections has created the need for object labeling in volume data sets for 3D visualization. A sequential, slice-to-slice approach is proposed that is less computational and memory intensive than 3D connected-component labeling while achieving better results than the 2D overlap sequential processing approach. Labeling occurs while tracking each object through the 3D volume via updated trajectory approximation.
This thesis is motivated by the desire to develop a labeling technique that captures key aspects of the human visual approach to the task. The proposed approach, while labeling 3D data, also provides a computationally efficient method by sequential processing of 2D slices instead of the whole 3D volume at once. Additionally, the proposed trajectory tracking approach performs correctly in many cases where current 2D sequential labeling techniques fail.
Trajectory tracking for labeling is a new approach, representing the 3D objects as curves and performing 3D curve tracing to label the approximate trajectories of the objects. The labeled trajectories are then mapped back to the 3D objects to complete the labeling process. Development of the proposed labeling approach is discussed while multiple examples are presented. These examples are used to illustrate that the proposed approach performs correctly where the current overlap approach fails; examples are also used to show that the behavior of the proposed approach parallels that of the typical human approach to object tracking.
Library of Congress Subject Headings
Tomography; Three-dimensional imaging in medicine; Diagnostic imaging; Image processing--Digital techniques
Electrical Engineering (MS)
Department, Program, or Center
Electrical Engineering (KGCOE)
S. A. Dianat
Vincent J. Amuso
Mikulis, Brandon Thomas, "Sequential slice object labeling in tomographic data via trajectory estimation" (2006). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus