The University of Southampton
University of Southampton Institutional Repository

Feature extraction in volumetric images

Feature extraction in volumetric images
Feature extraction in volumetric images
The increased interest in volumetric images in recent years requires new feature extraction methods for 3D image interpretation. The aim of this study is to provide algorithms that aid the process of detecting and segmenting geometrical objects from volumetric images. Due to high computational expense, such methods have yet to be established in the volumetric space. Only few have tackled this problem using shape descriptors and key-points of a specific shape; those techniques can detect complex shapes rather than simple geometric shapes due to the well defined key-points.

Simplifying the data in the volumetric image using a surface detector and surface curvature estimation preserves the important information about the shapes at the same time reducing the computational expense. Whilst the literature describes only the template of the three-dimensional Sobel operator and not its basis, we present an extended version of the Sobel operator, which considers the gradients of all directions to extract an object’s surface, and with clear basis that allows for development of larger operators. Surface curvature descriptors are usually based on geometrical properties of a segmented object rather than on the change in image intensity. In this work, a new approach is described to estimate the surface curvature of objects using local changes of image intensity. The new methods have shown reliable results on both synthetic and on real volumetric images.

The curvature and edge data are then processed in two new techniques for evidence gathering to extract a geometrical shape’s main axis or centre point. The accumulated data are taken directly from voxels’ geometrical locations rather than the surface normals as proposed in literature. The new approaches have been applied to detect a cylinder’s axis and spherical shapes. A new 3D line detection based on origin shifting has also been introduced. Accumulating, at every voxel, the angles resulting from a coordinate transform of a Cartesian to spherical system successfully indicates the existence of a 3D line in the volumetric image.

A novel method based on using an analogy to pressure is introduced to allow analysis/ visualisation of objects as though they have been separated, when they were actually touching in the original volumetric images. The approach provides a new domain highlighting the connected areas between multiple touching objects. A mask is formed to detach the interconnected objects and remarkable results are achieved. This is applied successfully to isolate coins within an image of a Roman hoard of coins, and other objects. The approach can fail to isolate objects when the space between them appears to be of similar density to the objects themselves. This motivated development of an operator extended by high-pass filtering and morphological operations. This led to more accurate extraction of coins within the Roman hoard, and to successful isolation of femurs in a database of scanned body images enabling better isolation of hip components in replacement therapy.
Alathari, Thamer
0e4d164a-4929-4f9c-8da0-31f525be5bcb
Alathari, Thamer
0e4d164a-4929-4f9c-8da0-31f525be5bcb
Nixon, Mark
2b5b9804-5a81-462a-82e6-92ee5fa74e12

Alathari, Thamer (2015) Feature extraction in volumetric images. University of Southampton, Physical Sciences and Engineering, Doctoral Thesis, 159pp.

Record type: Thesis (Doctoral)

Abstract

The increased interest in volumetric images in recent years requires new feature extraction methods for 3D image interpretation. The aim of this study is to provide algorithms that aid the process of detecting and segmenting geometrical objects from volumetric images. Due to high computational expense, such methods have yet to be established in the volumetric space. Only few have tackled this problem using shape descriptors and key-points of a specific shape; those techniques can detect complex shapes rather than simple geometric shapes due to the well defined key-points.

Simplifying the data in the volumetric image using a surface detector and surface curvature estimation preserves the important information about the shapes at the same time reducing the computational expense. Whilst the literature describes only the template of the three-dimensional Sobel operator and not its basis, we present an extended version of the Sobel operator, which considers the gradients of all directions to extract an object’s surface, and with clear basis that allows for development of larger operators. Surface curvature descriptors are usually based on geometrical properties of a segmented object rather than on the change in image intensity. In this work, a new approach is described to estimate the surface curvature of objects using local changes of image intensity. The new methods have shown reliable results on both synthetic and on real volumetric images.

The curvature and edge data are then processed in two new techniques for evidence gathering to extract a geometrical shape’s main axis or centre point. The accumulated data are taken directly from voxels’ geometrical locations rather than the surface normals as proposed in literature. The new approaches have been applied to detect a cylinder’s axis and spherical shapes. A new 3D line detection based on origin shifting has also been introduced. Accumulating, at every voxel, the angles resulting from a coordinate transform of a Cartesian to spherical system successfully indicates the existence of a 3D line in the volumetric image.

A novel method based on using an analogy to pressure is introduced to allow analysis/ visualisation of objects as though they have been separated, when they were actually touching in the original volumetric images. The approach provides a new domain highlighting the connected areas between multiple touching objects. A mask is formed to detach the interconnected objects and remarkable results are achieved. This is applied successfully to isolate coins within an image of a Roman hoard of coins, and other objects. The approach can fail to isolate objects when the space between them appears to be of similar density to the objects themselves. This motivated development of an operator extended by high-pass filtering and morphological operations. This led to more accurate extraction of coins within the Roman hoard, and to successful isolation of femurs in a database of scanned body images enabling better isolation of hip components in replacement therapy.

PDF
__soton.ac.uk_ude_personalfiles_users_jo1d13_mydesktop_Alathari.pdf - Other
Download (7MB)

More information

Published date: April 2015
Organisations: University of Southampton, Vision, Learning and Control

Identifiers

Local EPrints ID: 379936
URI: https://eprints.soton.ac.uk/id/eprint/379936
PURE UUID: 2dba4680-2eb6-4656-a6eb-aeb6cbbd8012
ORCID for Mark Nixon: ORCID iD orcid.org/0000-0002-9174-5934

Catalogue record

Date deposited: 18 Aug 2015 13:01
Last modified: 06 Jun 2018 13:18

Export record

Contributors

Author: Thamer Alathari
Thesis advisor: Mark Nixon ORCID iD

University divisions

Download statistics

Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.

View more statistics

Atom RSS 1.0 RSS 2.0

Contact ePrints Soton: eprints@soton.ac.uk

ePrints Soton supports OAI 2.0 with a base URL of https://eprints.soton.ac.uk/cgi/oai2

This repository has been built using EPrints software, developed at the University of Southampton, but available to everyone to use.

We use cookies to ensure that we give you the best experience on our website. If you continue without changing your settings, we will assume that you are happy to receive cookies on the University of Southampton website.

×