How well do computational features perceptually rank textures? A comparative evaluation

Xinghui Dong; Thomas S. Methven; Mike J. Chantler

doi:10.1145/2578726.2578762

How well do computational features perceptually rank textures? A comparative evaluation

Xinghui Dong, Thomas S. Methven, Mike J. Chantler

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

13 Citations (Scopus)

Abstract

Inspired by studies [4, 23, 40] which compared rankings obtained by search engines and human observers, in this paper we compare texture rankings derived by 51 sets of computational features against perceptual texture rankings obtained from a free-grouping experiment with 30 human observers, using a unify evaluation framework. Experimental results show that the MRSAR [37], VZNEIGHBORHOOD [62], LBPHF [2] and LBPBASIC [3] feature sets perform better than their counterparts. However, none of those feature sets are ideal. The best average G and M measures (measures of ranking accuracy from 0 to 1) [15, 5] obtained are 0.36 and 0.25 respectively. We suggest that this poor performance may be due to the small local neighborhood used to calculate higher-order features which cannot capture the long-range interactions that humans have been shown to exploit [14, 16, 49, 56]. Copyright is held by the owner/author(s).

Original language	English
Title of host publication	ICMR 2014 - Proceedings of the ACM International Conference on Multimedia Retrieval 2014
Publisher	Association for Computing Machinery
Pages	281-288
Number of pages	8
DOIs	https://doi.org/10.1145/2578726.2578762
Publication status	Published - 1 Jan 2014
Event	2014 4th ACM International Conference on Multimedia Retrieval - Glasgow, United Kingdom Duration: 1 Apr 2014 → 4 Apr 2014

Conference

Conference	2014 4th ACM International Conference on Multimedia Retrieval
Abbreviated title	ICMR 2014
Country/Territory	United Kingdom
City	Glasgow
Period	1/04/14 → 4/04/14

Keywords

Computational features
Evaluation
Perceptual texture ranking
Texture ranking
Texture retrieval
Texture similarity

ASJC Scopus subject areas

Computer Graphics and Computer-Aided Design
Human-Computer Interaction
Software

Access to Document

10.1145/2578726.2578762

Cite this

@inproceedings{bb786d700d8a493a9b25c3145dadcc00,

title = "How well do computational features perceptually rank textures? A comparative evaluation",

abstract = "Inspired by studies [4, 23, 40] which compared rankings obtained by search engines and human observers, in this paper we compare texture rankings derived by 51 sets of computational features against perceptual texture rankings obtained from a free-grouping experiment with 30 human observers, using a unify evaluation framework. Experimental results show that the MRSAR [37], VZNEIGHBORHOOD [62], LBPHF [2] and LBPBASIC [3] feature sets perform better than their counterparts. However, none of those feature sets are ideal. The best average G and M measures (measures of ranking accuracy from 0 to 1) [15, 5] obtained are 0.36 and 0.25 respectively. We suggest that this poor performance may be due to the small local neighborhood used to calculate higher-order features which cannot capture the long-range interactions that humans have been shown to exploit [14, 16, 49, 56]. Copyright is held by the owner/author(s).",

keywords = "Computational features, Evaluation, Perceptual texture ranking, Texture ranking, Texture retrieval, Texture similarity",

author = "Xinghui Dong and Methven, \{Thomas S.\} and Chantler, \{Mike J.\}",

year = "2014",

month = jan,

day = "1",

doi = "10.1145/2578726.2578762",

language = "English",

pages = "281--288",

booktitle = "ICMR 2014 - Proceedings of the ACM International Conference on Multimedia Retrieval 2014",

publisher = "Association for Computing Machinery",

address = "United States",

note = "2014 4th ACM International Conference on Multimedia Retrieval, ICMR 2014 ; Conference date: 01-04-2014 Through 04-04-2014",

}

Dong, X, Methven, TS & Chantler, MJ 2014, How well do computational features perceptually rank textures? A comparative evaluation. in ICMR 2014 - Proceedings of the ACM International Conference on Multimedia Retrieval 2014. Association for Computing Machinery, pp. 281-288, 2014 4th ACM International Conference on Multimedia Retrieval, Glasgow, United Kingdom, 1/04/14. https://doi.org/10.1145/2578726.2578762

How well do computational features perceptually rank textures? A comparative evaluation. / Dong, Xinghui; Methven, Thomas S.; Chantler, Mike J.
ICMR 2014 - Proceedings of the ACM International Conference on Multimedia Retrieval 2014. Association for Computing Machinery, 2014. p. 281-288.

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

TY - GEN

T1 - How well do computational features perceptually rank textures? A comparative evaluation

AU - Dong, Xinghui

AU - Methven, Thomas S.

AU - Chantler, Mike J.

PY - 2014/1/1

Y1 - 2014/1/1

N2 - Inspired by studies [4, 23, 40] which compared rankings obtained by search engines and human observers, in this paper we compare texture rankings derived by 51 sets of computational features against perceptual texture rankings obtained from a free-grouping experiment with 30 human observers, using a unify evaluation framework. Experimental results show that the MRSAR [37], VZNEIGHBORHOOD [62], LBPHF [2] and LBPBASIC [3] feature sets perform better than their counterparts. However, none of those feature sets are ideal. The best average G and M measures (measures of ranking accuracy from 0 to 1) [15, 5] obtained are 0.36 and 0.25 respectively. We suggest that this poor performance may be due to the small local neighborhood used to calculate higher-order features which cannot capture the long-range interactions that humans have been shown to exploit [14, 16, 49, 56]. Copyright is held by the owner/author(s).

AB - Inspired by studies [4, 23, 40] which compared rankings obtained by search engines and human observers, in this paper we compare texture rankings derived by 51 sets of computational features against perceptual texture rankings obtained from a free-grouping experiment with 30 human observers, using a unify evaluation framework. Experimental results show that the MRSAR [37], VZNEIGHBORHOOD [62], LBPHF [2] and LBPBASIC [3] feature sets perform better than their counterparts. However, none of those feature sets are ideal. The best average G and M measures (measures of ranking accuracy from 0 to 1) [15, 5] obtained are 0.36 and 0.25 respectively. We suggest that this poor performance may be due to the small local neighborhood used to calculate higher-order features which cannot capture the long-range interactions that humans have been shown to exploit [14, 16, 49, 56]. Copyright is held by the owner/author(s).

KW - Computational features

KW - Evaluation

KW - Perceptual texture ranking

KW - Texture ranking

KW - Texture retrieval

KW - Texture similarity

U2 - 10.1145/2578726.2578762

DO - 10.1145/2578726.2578762

M3 - Conference contribution

AN - SCOPUS:84899728375

SP - 281

EP - 288

BT - ICMR 2014 - Proceedings of the ACM International Conference on Multimedia Retrieval 2014

PB - Association for Computing Machinery

T2 - 2014 4th ACM International Conference on Multimedia Retrieval

Y2 - 1 April 2014 through 4 April 2014

ER -

How well do computational features perceptually rank textures? A comparative evaluation

Abstract

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Fingerprint

Cite this