|
Projects
Mediana
Duration
2000 - 2001
Summary
A computer-based workplace for media researchers is developed which consists
of tools facilitating the management and analysis of media data (i.e. text,
image, video documents). Many tools are specifically designed for media
researcher in order to precisely map their way of working and analysis.
The workspace mainly consists of:
-) data and media management tools in order to manage public and private media
documents, their attribute data and desciptive meta data;
-) automatic image and video analysis tools in order to generate relevant
information about their structure and content (three analysis tools are
developed: cut/shot detection, text detection and face
detection); and a
-) graphical user interface which supports editing, retrieval and visualization
tools for all kind of data used by the media researcher. An important feature
is the visualization of relationship data existing between different data
and media objects.
A prototype of this workplace is already in use.
Homepage
Project Z2: Methods and Tools for Computer-based Media Research Analysis
of Audiovisual Information
Documents
[LuttermannFreisleben02]
VRML Wavelet
Duration
1998- 2000
Summary
New compressing techniques are developed which allows to load progressively
temporal graphic objects, e.g. temporal meshes, over the internet for visualization
on the local display. In the first step only a crude model of the temporal
graphic object is transmitted, however, which is good for navigation and
first analysis. Then, the model will be refined by continuously loading
additional information until the complete model is transmitted. The techniques
are based on progressive meshes and the wavelet transformation.
Documents
[Luttermann03]
VRML History
Duration
1997- 2000
Summary
The standard file format and graphic language VRML is extended with new
language concepts which allow an author to describe temporal graphical
objects and time values based on valid time (which is different from already
supported animation/presentation time). Additionally, a browser is extended
which reads, interprets and visualizes this new type of content.
Homepage
VRML History Homepage
Documents
[LuttermannLangner00]
[LuttermannGrauer99b]
[LuttermannGrauer99a]
[LuttermannGrauer98b]
ST-EER Model
Duration
1995- 1996
Summary
The classical Enhanced Entity-Relationship (EER) model, which is used to
describe a conceptual view of the application's data structure, is extended
with concepts for the description of spatial and temporal aspects of data
structures.
Documents
[Luttermann96]
Chronos
Duration
1993- 1995
Summary
Tools and techniques for the management and visualization of temporal GIS-data
are developed. An implementation of these new tools is integrated in the
commercial geographic information system GRADIS-GIS.
Homepage
Project CHRONOS
Documents
[LuttermannBlobel95b]
[BlobelLuttermann95]
[LuttermannBlobel95a]
[Luttermann94]
Faceman
Duration
1991- 1992
Summary
A face recognition software is developed based on the Principal Component
Analysis (PCA) and a back-propagation network. In the initialisation phase
the system is trained with a set of "known" faces of some persons. Then
faces unknown to the system are presented to the system to be detected
and assigned to one person. In test runs with 100 faces (selected from
the set of faces of 20 persons) a detection rate of 85 % is reached.
Each face recorded in an image is compressed by the PCA and stored
as a coordinate vector of a new normalized coordinate system of eigenvectors
(the principal components) determined by the set of known faces. The eigenvectors
are called eigenfaces since they are similar to faces. The coordinate vector
of each face is then presented to a back-propagation network which is trained
by the coordinate vectors of the set of known faces.
|
Documents
[GroßLuttermann93]
Go-Moku-Net
Duration
1990- 1991
Summary
A game program based on neural networks is developed which is used to play
the board game Go-Moku (well-known by the name "5-wins"). The software
analyses the games move by move, trains the integrated back-propagation
network and thus tries to improve its playing technique.
|
Documents
[FreislebenLuttermann96]
last updated at 2003-09-13, H. Luttermann (mail to hartmut@HartmutLuttermann.de) |
|