All technical report files are in compressed (gzip) postscript format.
The file names are of the sort BU-CE-YYXX.pdf

<br><hr><br> 
<A href="http://www.cs.bilkent.edu.tr/tech-reports/2006/BU-CE-0601.pdf" 
target=_blank>BU-CE-0601</A><br><h3>
TITLE: 
On the Convergence of a Class of Multilevel Methods for Large, Sparse
Markov Chains

<br>
AUTHORS: Peter Buchholz, Tugrul Dayar
</h3>

ABSTRACT:
This paper investigates the theory behind the steady state analysis of
large, sparse Markov chains (MCs) with a recently proposed class of
multilevel (ML) methods using concepts from algebraic multigrid and
iterative aggregation-disaggregation. The motivation is to better
understand the convergence characteristics of the class of ML methods
and to have a clearer formulation that will aid their
implementation. In doing this, restriction (or aggregation) and
prolongation (or disaggregation) operators of multigrid are used, and
the Kronecker based approach for hierarchical Markovian models (HMMs)
is employed, since it suggests a natural and compact definition of
grids (or levels). However, the HMM formalism used to describe the
class of ML methods for large, sparse MCs has no influence on the
theoretical results derived. 

<P><b>Keywords:</b>
Markov chains, multilevel methods, multigrid,
aggregation-disaggregation, Kronecker based numerical techniques

<br><hr><br>
<A href="http://www.cs.bilkent.edu.tr/tech-reports/2006/BU-CE-0602.pdf" 
target=_blank>BU-CE-0602</A><br><h3>
TITLE: 
Conservative Occlusion Culling for Urban Visualization using a
Slice-wise Data Structure

<br>
AUTHORS: Turker Yilmaz and Ugur Gudukbay
</h3>

ABSTRACT:
In this paper, we propose a new shrinking process for conservative
from-point  occlusion culling algorithms and a data structure for the
visualization of urban environments.
The visible geometry in a typical urban walkthrough mainly consists of
partially visible buildings. Occlusion-culling algorithms, in which
the granularity is based on buildings, render these partially visible
buildings completely.
We observe that the visibility in urban walkthroughs shows certain
characteristics. The proposed \textit {slice-wise data structure}
represents the buildings, exploiting these characteristics, in terms
of slices parallel to the coordinate axes. This forms the base for
occlusion culling where the occlusion granularity is at slice
level. The proposed slice-wise data structure has minimal storage
requirements. The visible parts can be accessed at constant time
during navigation  with the help of a preprocessing stage. We also
propose to shrink the occluders in a scene. This is necessary for a
conservative from-point occlusion culling algorithm, which can also be
applied to nonconvex general 3D occluders.
Empirical results show that
a 54\% decrease in the number of processed polygons and 46\% speed-up
in frame-rate can be achieved by using the proposed conservative
occlusion-culling algorithm with rather than an occlusion-culling
method  where the granularity is based on individual buildings.

<P><b>Keywords:</b>
octree, occluder shrinking, from-point occlusion culling, urban
visualization, visibility preprocessing.

<br><hr><br>

<A href="http://www.cs.bilkent.edu.tr/tech-reports/2006/BU-CE-0603.pdf" 
target=_blank>BU-CE-0603</A><br><h3>
TITLE: 
Motion Capture from Single Video Sequence

<br>
AUTHORS: Ibrahim Demir
</h3>

ABSTRACT:
3D human pose reconstruction is a popular research area since it can
be used in various applications. Currently most of the methods work
for constrained environments, where multi camera views are available
and camera calibration is known, or a single camera view is available,
which requires intensive user effort.
However most of the currently available data do not satisfy these
constraints, thus they cannot be processed by these algorithms. In
this thesis a framework is proposed to reconstruct 3D pose of a human
for animation from a sequence of single view video frames. The
framework for pose construction starts with background
estimation. Once the image background is estimated, the body
silhouette is extracted by using image subtraction for each
frame. Then the body silhouettes are automatically labeled by using a
model-based approach. Finally, the 3D pose is constructed from the
labeled human silhouette by assuming orthographic projection. The
proposed approach does not require camera calibration.
The proposed framework assumes that the input video has a static
background and it has no significant perspective effects and the
performer is in upright position.
<P><b>Keywords:</b>
motion capture, framework, single camera, uncalibrated camera,
vision-based, animation.

<br><hr><br>