Online figure-ground segmentation with edge pixel classification

Zhaozheng Yin, Robert T. Collins

Research output: Contribution to conferencePaper

13 Citations (Scopus)

Abstract

The need for figure-ground segmentation in video arises in many vision problems like tracker initialization, accurate object shape representation and drift-free appearance model adaptation. This paper uses a 3D spatio-temporal Conditional Random Field (CRF) to combine different segmentation cues while enforcing temporal coherence. Without supervised parameter training, the weighting factors for different data potential functions in the CRF model are adapted online to reflect changes in object appearance and environment. To get an accurate boundary based on the 3D CRF segmentation result, edge pixels are classified into three classes: foreground, background and boundary. The final foreground region bitmask is constructed from the foreground and boundary edge pixels. The effectiveness of our approach is demonstrated on several airborne videos with large appearance change and heavy occlusion.

Original languageEnglish (US)
DOIs
StatePublished - Jan 1 2008
Event2008 19th British Machine Vision Conference, BMVC 2008 - Leeds, United Kingdom
Duration: Sep 1 2008Sep 4 2008

Other

Other2008 19th British Machine Vision Conference, BMVC 2008
CountryUnited Kingdom
CityLeeds
Period9/1/089/4/08

Fingerprint

Pixels

All Science Journal Classification (ASJC) codes

  • Computer Vision and Pattern Recognition

Cite this

Yin, Z., & Collins, R. T. (2008). Online figure-ground segmentation with edge pixel classification. Paper presented at 2008 19th British Machine Vision Conference, BMVC 2008, Leeds, United Kingdom. https://doi.org/10.5244/C.22.35
Yin, Zhaozheng ; Collins, Robert T. / Online figure-ground segmentation with edge pixel classification. Paper presented at 2008 19th British Machine Vision Conference, BMVC 2008, Leeds, United Kingdom.
@conference{44ce85a59506400e88912bf48978e735,
title = "Online figure-ground segmentation with edge pixel classification",
abstract = "The need for figure-ground segmentation in video arises in many vision problems like tracker initialization, accurate object shape representation and drift-free appearance model adaptation. This paper uses a 3D spatio-temporal Conditional Random Field (CRF) to combine different segmentation cues while enforcing temporal coherence. Without supervised parameter training, the weighting factors for different data potential functions in the CRF model are adapted online to reflect changes in object appearance and environment. To get an accurate boundary based on the 3D CRF segmentation result, edge pixels are classified into three classes: foreground, background and boundary. The final foreground region bitmask is constructed from the foreground and boundary edge pixels. The effectiveness of our approach is demonstrated on several airborne videos with large appearance change and heavy occlusion.",
author = "Zhaozheng Yin and Collins, {Robert T.}",
year = "2008",
month = "1",
day = "1",
doi = "10.5244/C.22.35",
language = "English (US)",
note = "2008 19th British Machine Vision Conference, BMVC 2008 ; Conference date: 01-09-2008 Through 04-09-2008",

}

Yin, Z & Collins, RT 2008, 'Online figure-ground segmentation with edge pixel classification' Paper presented at 2008 19th British Machine Vision Conference, BMVC 2008, Leeds, United Kingdom, 9/1/08 - 9/4/08, . https://doi.org/10.5244/C.22.35

Online figure-ground segmentation with edge pixel classification. / Yin, Zhaozheng; Collins, Robert T.

2008. Paper presented at 2008 19th British Machine Vision Conference, BMVC 2008, Leeds, United Kingdom.

Research output: Contribution to conferencePaper

TY - CONF

T1 - Online figure-ground segmentation with edge pixel classification

AU - Yin, Zhaozheng

AU - Collins, Robert T.

PY - 2008/1/1

Y1 - 2008/1/1

N2 - The need for figure-ground segmentation in video arises in many vision problems like tracker initialization, accurate object shape representation and drift-free appearance model adaptation. This paper uses a 3D spatio-temporal Conditional Random Field (CRF) to combine different segmentation cues while enforcing temporal coherence. Without supervised parameter training, the weighting factors for different data potential functions in the CRF model are adapted online to reflect changes in object appearance and environment. To get an accurate boundary based on the 3D CRF segmentation result, edge pixels are classified into three classes: foreground, background and boundary. The final foreground region bitmask is constructed from the foreground and boundary edge pixels. The effectiveness of our approach is demonstrated on several airborne videos with large appearance change and heavy occlusion.

AB - The need for figure-ground segmentation in video arises in many vision problems like tracker initialization, accurate object shape representation and drift-free appearance model adaptation. This paper uses a 3D spatio-temporal Conditional Random Field (CRF) to combine different segmentation cues while enforcing temporal coherence. Without supervised parameter training, the weighting factors for different data potential functions in the CRF model are adapted online to reflect changes in object appearance and environment. To get an accurate boundary based on the 3D CRF segmentation result, edge pixels are classified into three classes: foreground, background and boundary. The final foreground region bitmask is constructed from the foreground and boundary edge pixels. The effectiveness of our approach is demonstrated on several airborne videos with large appearance change and heavy occlusion.

UR - http://www.scopus.com/inward/record.url?scp=84898472695&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84898472695&partnerID=8YFLogxK

U2 - 10.5244/C.22.35

DO - 10.5244/C.22.35

M3 - Paper

AN - SCOPUS:84898472695

ER -

Yin Z, Collins RT. Online figure-ground segmentation with edge pixel classification. 2008. Paper presented at 2008 19th British Machine Vision Conference, BMVC 2008, Leeds, United Kingdom. https://doi.org/10.5244/C.22.35