Parallel Mean Shift Accuracy and Performance Trade-Offs

Kirsty Duncan; Robert James Stewart; Gregory John Michaelson

doi:10.1109/ICIP.2018.8451199

Parallel Mean Shift Accuracy and Performance Trade-Offs

Kirsty Duncan, Robert James Stewart, Gregory John Michaelson

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

44 Downloads (Pure)

Abstract

This paper decomposes the algorithmic parameters that affect the accuracy and parallel run times of mean shift segmentation. Following Comaniciu and Meer [1], rather than perform calculations in the feature space of the image, the joint spatial-range domain is represented by the image space, with feature space information associated with each point. We report parallel speedup and segmentation accuracy using a standardised segmentation dataset and the Probabilistic Rand index (PRI) accuracy measure. Changes to the algorithmic parameters are analysed and a sweet spot between PRI and run time is found. Using a range window radius of 20, spatial window radius of 10 and threshold of 50, the PRI is improved by 0.17, an increase of 34% which is comparable to state of the art. Mean shift clustering run time is reduced by 97% with parallelism, a speedup of 32 on a 64-core CPU.

Original language	English
Title of host publication	2018 25th IEEE International Conference on Image Processing (ICIP)
Publisher	IEEE
Pages	2197-2201
Number of pages	5
ISBN (Electronic)	9781479970612
DOIs	https://doi.org/10.1109/ICIP.2018.8451199
Publication status	Published - 6 Sept 2018
Event	25th IEEE International Conference on Image Processing 2018 - Athens, Greece Duration: 7 Oct 2018 → 10 Oct 2018 https://2018.ieeeicip.org/

Conference

Conference	25th IEEE International Conference on Image Processing 2018
Abbreviated title	IEEE ICIP 2018
Country/Territory	Greece
City	Athens
Period	7/10/18 → 10/10/18
Internet address	https://2018.ieeeicip.org/

Access to Document

10.1109/ICIP.2018.8451199

Download pdf
© 2018 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.
Accepted author manuscript, 902 KBLicence: All Rights Reserved

Open dataset for "Parallel Mean Shift Accuracy and Performance Trade-Offs"
Stewart, R. J. (Contributor), Duncan, K. (Creator) & Michaelson, G. J. (Contributor), Heriot-Watt University, 24 May 2018
DOI: 10.17861/fd3ee9dd-dd6d-47f2-8b22-12e634cb6556
Dataset

Cite this

@inproceedings{fd638e780d944d64ae53c2a14a93afcd,

title = "Parallel Mean Shift Accuracy and Performance Trade-Offs",

abstract = "This paper decomposes the algorithmic parameters that affect the accuracy and parallel run times of mean shift segmentation. Following Comaniciu and Meer [1], rather than perform calculations in the feature space of the image, the joint spatial-range domain is represented by the image space, with feature space information associated with each point. We report parallel speedup and segmentation accuracy using a standardised segmentation dataset and the Probabilistic Rand index (PRI) accuracy measure. Changes to the algorithmic parameters are analysed and a sweet spot between PRI and run time is found. Using a range window radius of 20, spatial window radius of 10 and threshold of 50, the PRI is improved by 0.17, an increase of 34% which is comparable to state of the art. Mean shift clustering run time is reduced by 97% with parallelism, a speedup of 32 on a 64-core CPU.",

author = "Kirsty Duncan and Stewart, {Robert James} and Michaelson, {Gregory John}",

year = "2018",

month = sep,

day = "6",

doi = "10.1109/ICIP.2018.8451199",

language = "English",

pages = "2197--2201",

booktitle = "2018 25th IEEE International Conference on Image Processing (ICIP)",

publisher = "IEEE",

address = "United States",

note = "25th IEEE International Conference on Image Processing 2018, IEEE ICIP 2018 ; Conference date: 07-10-2018 Through 10-10-2018",

url = "https://2018.ieeeicip.org/",

}

TY - GEN

T1 - Parallel Mean Shift Accuracy and Performance Trade-Offs

AU - Duncan, Kirsty

AU - Stewart, Robert James

AU - Michaelson, Gregory John

PY - 2018/9/6

Y1 - 2018/9/6

N2 - This paper decomposes the algorithmic parameters that affect the accuracy and parallel run times of mean shift segmentation. Following Comaniciu and Meer [1], rather than perform calculations in the feature space of the image, the joint spatial-range domain is represented by the image space, with feature space information associated with each point. We report parallel speedup and segmentation accuracy using a standardised segmentation dataset and the Probabilistic Rand index (PRI) accuracy measure. Changes to the algorithmic parameters are analysed and a sweet spot between PRI and run time is found. Using a range window radius of 20, spatial window radius of 10 and threshold of 50, the PRI is improved by 0.17, an increase of 34% which is comparable to state of the art. Mean shift clustering run time is reduced by 97% with parallelism, a speedup of 32 on a 64-core CPU.

AB - This paper decomposes the algorithmic parameters that affect the accuracy and parallel run times of mean shift segmentation. Following Comaniciu and Meer [1], rather than perform calculations in the feature space of the image, the joint spatial-range domain is represented by the image space, with feature space information associated with each point. We report parallel speedup and segmentation accuracy using a standardised segmentation dataset and the Probabilistic Rand index (PRI) accuracy measure. Changes to the algorithmic parameters are analysed and a sweet spot between PRI and run time is found. Using a range window radius of 20, spatial window radius of 10 and threshold of 50, the PRI is improved by 0.17, an increase of 34% which is comparable to state of the art. Mean shift clustering run time is reduced by 97% with parallelism, a speedup of 32 on a 64-core CPU.

U2 - 10.1109/ICIP.2018.8451199

DO - 10.1109/ICIP.2018.8451199

M3 - Conference contribution

SP - 2197

EP - 2201

BT - 2018 25th IEEE International Conference on Image Processing (ICIP)

PB - IEEE

T2 - 25th IEEE International Conference on Image Processing 2018

Y2 - 7 October 2018 through 10 October 2018

ER -

Parallel Mean Shift Accuracy and Performance Trade-Offs

Abstract

Conference

Access to Document

Fingerprint

Datasets

Open dataset for "Parallel Mean Shift Accuracy and Performance Trade-Offs"

Cite this