Optimum slice reduction algorithm for fast surface reconstruction from contour slices
Loading...
Date
2007
Authors
Tang, Sze Ling
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This thesis is concerned with the reconstruction of surface from a series of
contour slices, with the aim to speed up the reconstruction process while preserving
the output quality at an acceptable level. The proposed technique in this thesis, preprocesses
the slices of contour prior to surface reconstruction. This shift of focus to the
input seems reasonable for several reasons. First, modern imaging modalities are built
to capture high precision data, some of these are redundant or visually insignificant.
Second, the dataset’s size often determines the computational time of the
reconstruction process hence if we provide with lesser but significant number of slices,
there is a chance that we can improve the speed of the reconstruction process. Third,
majority of the existing works focus either on the reconstruction algorithm itself or on
the post-processing of the resulting meshes, the method proposed here provides an
alternative way to address the issue between speed of computation and accuracy of
output.
Our proposed method is divided into two phases: Optimum Slice Reduction and
Block Triangulation. In the first phase, redundant slices are reduced based on the
number of contour point’s differences for a pair of adjacent slices. For any given
dataset, an optimum threshold value is extracted, this value is used to determine the
slices to be removed (or retained). Once this process is completed, a set of nonconsecutive
slices is produced. This requires special reconstruction method to rebuild
surface(s) from sparse contour slices, and this problem is addressed in the second
phase of our algorithm. Our modified meshing algorithm is based on the VTK’s module
called vtkDelaunay3D. The crucial part in the modification is the automatic generations
of alpha value for different “blocks” of the sparse contour slices. For each block, we
derived a suitable alpha value, which we then applied the vtkDelaunay3D to
reconstruct the surface from the block of slices.
We assess the effectiveness of the proposed method through ten (10) datasets
with various types, complexities and sizes. For each dataset, we collect quantitative
measure like the reconstruction time and the qualitative measures via visual
comparison. Through the experiments, we achieved slice reduction in the range
between 40 – 60 % and this result has improved the reconstruction time between 60 -
90 %. Through visual comparison, we showed the regions in visual output that are
overlapping or detaching (between the original vs. reduced). These are consistent with
the principles set in our proposed method.
We conclude that the proposed method improves the computational speed of
the reconstruction process with an acceptable level of output quality; major
improvement is evident in the case of large and complex dataset.
Description
Master
Keywords
Computer science , Algorithm , Contour slices