Topological analysis of images

Matjaž Cerar (2015) Topological analysis of images. EngD thesis.

Preview

PDF Download (1005Kb)

Abstract

In this thesis we present an implementation of a topological approach to 2-dimensional digital images. First, we present two methods for simplifying and preparing the image, without loss of information, for further algorithms. We represent the image as a topological structure called a cubical complex. On the cubical complex, a discrete vector field encoding the directions of descent of grey scale values is constructed, together with the corresponding list of critical cells. From these, the Morse complex, which captures the vital information about the image, is built. Using Betti numbers, important features in the image are described. We present two approaches to computing Betti numbers. The thesis concludes with a presentation of how the implemented algorithms can be used for counting bright objects on specific examples of images.

Actions (login required)