Orientation and anisotropy of multi-component shapes

Speaker:        Dr. Jovisa Zunic
                University of Exeter
                U.K

Title:          "Orientation and anisotropy of multi-component shapes"

Date:           Monday, 16 September 2013

Time:           4:00pm - 5:00pm

Venue:          Lecture Theatre F (near lifts 25/26), HKUST

Abstract:

There are many situations in which several single objects are better
considered as components of a multi-component shape (e.g. a shoal of
fish), but there are also situations in which a single object is better
segmented into natural components and considered as a multi-component
shape (e.g. decomposition of cellular materials onto the corresponding
cells). Interestingly, not much research has been done on multi-component
shapes. Recently, the orientation and anisotropy problems were considered
and some solutions have been offered.

The object orientation problem is a recurrent problem in image processing
and computer vision. It is usually an initial step or a part of data
pre-processing, implying that an unsuitable solution could lead to a large
cumulative error at the end of the vision system's pipeline. We review the
new idea for the orientation of multi-component shapes, and also its
relation to the most standard method for determining the orientation of
single-component shapes. We also show how the anisotropy measure of
multi-component shapes, as a quantity which indicates how consistently the
shape components are oriented, can be obtained as a by-product of the
approach used.

********************
Biography:

Jovisa Zunic is a senior lecturer at the College of Engineering,
Mathematics and Physical Sciences, University of Exeter, U.K. He also
holds a professorship at the Mathematical Institute, Serbian Academy of
Sciences and Arts, Belgrade, Serbia. His recent research is mainly focused
on image processing, computer vision, and pattern recognition problems. A
particular attention is given to the shape based approaches. His research
also includes development of efficient digital object encoding schemes,
estimation of the features of real objects from the corresponding digital
images and limitations in such estimations. Most of those results are
published in the leading computer science journals, as they are:  Computer
Vision and Image Understanding, IEEE T-IP, IEEE T-IT, IEEE T-NN, IEEE
T-PAMI, IEEE T-PDS, International Journal of Computer Vision, Journal of
Mathematical Imaging and Vision, Pattern Recognition, SIAM Journal on
Imaging Sciences, etc.

Several new mathematical results had to be established in order to solve
some of the problems mentioned above. Most of them are related to the
squared (cubed) integer grids (what is actually a mathematical model for
2D and higher dimensional digital images). Results are presented in high
quality mathematical journals (e.g. Acta Arithmetica, Advances in Applied
Mathematics, Proceedings of LMS, Discrete Mathematics, Forum Mathematicum,
Foundations of Computational Mathematics, Journal of Combinatorial Theory
- A,  Journal of Number Theory, etc).