Research Interests
Back to Falcão's home page
Image Processing, Visualization, Manipulation, and Analysis
Images
have become one of the main sources of information to study
physical phenomena and design intelligent systems. In their
most general form, these images can be multi-dimensional (e.g., CT and
MR images), multi-spectral (e.g., remote sensing images), or both
(e.g., digital video). We are interested in image processing and
analysis
techniques that enhance, extract, represent, describe, and
classify the image content. Particularly, we have studied graph-based
image
operators (i.e., Image Foresting Transforms) for filtering,
segmentation, representation, description, detection, and recognition.
We
are also interested in techniques for visualization, manipulation,
quantitative analysis
of 3D biomedical structures for the purpose of diagnosis, treatment,
surgical planning, and
research.
Information Annotation, Organization, and Retrieval
Once
the data are extracted from the
images, they must be combined with other informations about the problem
and stored into a database to support different types of queries. We
are
interested in image annotation, organization, and retrieval techniques
to maintain
those databases. For instance, we have investigated active learning
techniques to annotate image databases and to reduce the semantic gap
between content-based image
representation and user's expectation in Content-Based Image Retrieval
(CBIR) systems. In CBIR, our aim has been to customize the information
system to
satisfy the expectations of a given user or user group, with a minimum
of user involvement.
Machine Learning and Pattern Recognition
For
a given database, we are interested in optimization techniques
for feature selection, distance combination, statistical data analysis,
clustering, and classification. We
have investigated, for instance, machine learning and pattern
recognition techniques based on optimum-path forests. These techniques
map the data into a graph in the feature space and exploit connectivity
functions for clustering and classification. Although our goal
has been automation in well controlled
applications, such as the diagnosis of parasites, we are also very
interested in minimizing user involvement in active learning
applications, where the concept of an image may be different for
distinct users.
Applications
The
applications in Geology,
Agriculture, Biology, Medicine, and Engineering have been very
important to inspire the development of
new methods in each of the aforementioned research topics. We
have been working on MR-image analysis of the brain, CT-image
segmentation of the thorax, tooth fracture detection in CBCT images,
the automation of the diagnosis of intestinal parasites using
microscopy images, face recognition systems, coffee crop segmentation
in remote sensing images, interactive image/video edition, image
segmentation of rock fragments, pose recognition in video, etc.