EVENT

Talk by Dr. Yukiko Kenmochi (CNRS Research Associate at Laboratoire d'Informatique Gaspard-Monge, Universite Paris-Est, France)
"Hierarchy processing with graphs"

Title:

Hierarchy processing with graphs

Speaker:

Dr. Yukiko Kenmochi (CNRS Research Associate at Laboratoire d'Informatique Gaspard-Monge, Universite Paris-Est, France)

Time/Date:

16:30-17:30　/　January 20th, Monday

Place:

20F meeting room #2005, NII
https://www.nii.ac.jp/en/about/access/

Abstract:

Hierarchical data analysis is useful and became very popular in image processing recently. The theory of hierarchical image representations have been well studied in Mathematical Morphology, and it provides a useful framework to treat objects or regions in digital images taking into account theirs scales and inclusion relations. In this talk, we first explain that such hierarchies can be represented by edge-weighted graphs. Thanks to this graphical framework for hierarchies, some hierarchical problems are re-written as re-weighting edge-weighted graphs if they can be considered as problems of transforming/modifying hierarchies. We show how to make such re-weighting efficiently for (1) hierarchizing image segmentation algorithms, and (2) removing non-significant regions in hierarchical segmentation/clustering.