Superpixel Finite Element Segmentation for RGB-D Images

Dermot Kerr; Sonya Coleman; Scotney Bryan

Superpixel Finite Element Segmentation for RGB-D Images

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

180 Downloads (Pure)

Abstract

Computer vision research has advanced from focusing solely on intensity images to the use of depth images, or combinations of RGB, intensity and depth images, mainly due to the recent development of low cost depth cameras. These images can be efficiently represented as a space-variant image by segmenting the images using a superpixel representation. Whilst superpixel representations offer advantages in terms of reduced processing requirements they present challenges in further processing as many existing image processing techniques require regularly distributed image data. We overcome this issue by making use of the Finite element framework for processing these images and demonstrate the application of the technique for detecting access holes in disaster management situations.

Original language	English
Title of host publication	Unknown Host Publication
Publisher	IEEE
Pages	1-5
Number of pages	5
Publication status	Accepted - 24 Apr 2017
Event	3rd International Conference on Robotics and Vision (ICRV 2017) - Wuhan, China Duration: 24 Apr 2017 → … http://www.icrv.org/

Conference

Conference	3rd International Conference on Robotics and Vision (ICRV 2017)
Period	24/04/17 → …
Internet address	http://www.icrv.org/

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 9 Industry, Innovation, and Infrastructure

Keywords

RGB-D imaging
image segmentation
SLIC
finite element framework
feature detection

Access to Document

RV18 acceptedAuthor Accepted Manuscript, 787 KB

Cite this

@inproceedings{fe98ce497bd549cdb7570714ec7f96dd,

title = "Superpixel Finite Element Segmentation for RGB-D Images",

abstract = "Computer vision research has advanced from focusing solely on intensity images to the use of depth images, or combinations of RGB, intensity and depth images, mainly due to the recent development of low cost depth cameras. These images can be efficiently represented as a space-variant image by segmenting the images using a superpixel representation. Whilst superpixel representations offer advantages in terms of reduced processing requirements they present challenges in further processing as many existing image processing techniques require regularly distributed image data. We overcome this issue by making use of the Finite element framework for processing these images and demonstrate the application of the technique for detecting access holes in disaster management situations.",

keywords = "RGB-D imaging, image segmentation, SLIC, finite element framework, feature detection",

author = "Dermot Kerr and Sonya Coleman and Scotney Bryan",

year = "2017",

month = apr,

day = "24",

language = "English",

pages = "1--5",

booktitle = "Unknown Host Publication",

publisher = "IEEE",

address = "United States",

note = "3rd International Conference on Robotics and Vision (ICRV 2017) ; Conference date: 24-04-2017",

url = "http://www.icrv.org/",

}

TY - GEN

T1 - Superpixel Finite Element Segmentation for RGB-D Images

AU - Kerr, Dermot

AU - Coleman, Sonya

AU - Bryan, Scotney

PY - 2017/4/24

Y1 - 2017/4/24

N2 - Computer vision research has advanced from focusing solely on intensity images to the use of depth images, or combinations of RGB, intensity and depth images, mainly due to the recent development of low cost depth cameras. These images can be efficiently represented as a space-variant image by segmenting the images using a superpixel representation. Whilst superpixel representations offer advantages in terms of reduced processing requirements they present challenges in further processing as many existing image processing techniques require regularly distributed image data. We overcome this issue by making use of the Finite element framework for processing these images and demonstrate the application of the technique for detecting access holes in disaster management situations.

AB - Computer vision research has advanced from focusing solely on intensity images to the use of depth images, or combinations of RGB, intensity and depth images, mainly due to the recent development of low cost depth cameras. These images can be efficiently represented as a space-variant image by segmenting the images using a superpixel representation. Whilst superpixel representations offer advantages in terms of reduced processing requirements they present challenges in further processing as many existing image processing techniques require regularly distributed image data. We overcome this issue by making use of the Finite element framework for processing these images and demonstrate the application of the technique for detecting access holes in disaster management situations.

KW - RGB-D imaging

KW - image segmentation

KW - SLIC

KW - finite element framework

KW - feature detection

M3 - Conference contribution

SP - 1

EP - 5

BT - Unknown Host Publication

PB - IEEE

T2 - 3rd International Conference on Robotics and Vision (ICRV 2017)

Y2 - 24 April 2017

ER -

Superpixel Finite Element Segmentation for RGB-D Images

Abstract

Conference

UN SDGs

Keywords

Access to Document

Fingerprint

Cite this