Presentation 2001/5/11
Quantitative Evaluation of Segmentation Results of Color Images
Masaki MATSUURA, Keiko TAKAHASHI, Takahiro SUGIYAMA, Keiichi ABE,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this paper we report a study on the evaluation of segmentation results of color images. First we made human evaluation experiments for evaluating various segmentaion results of eleven color images, using ten-scale grading, by eight subjects. Next we proposed three criteria for evaluating segmentation results: (1) uniformity of colors in a region, (2) number of proper region boundary pixels, and (3) number of improper region boundary pixels, and their evaluating formulas. In order to integrate the above three evaluations into a single evaluating function, we made a regression analysis from the three kinds of evaluating values to the average human evaluated scores of the segmentation results. If we omit undersegmented results, we obtained an evaluating function producing estimations satisfactorily close to human evaluations, both for training images and for images not used in the regression analysis.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Color image / Segmentation / Segmentation results / Performance evaluation / Uniformity of region / Region boundary
Paper # PRMU2001-15
Date of Issue

Conference Information
Committee PRMU
Conference Date 2001/5/11(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Pattern Recognition and Media Understanding (PRMU)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Quantitative Evaluation of Segmentation Results of Color Images
Sub Title (in English)
Keyword(1) Color image
Keyword(2) Segmentation
Keyword(3) Segmentation results
Keyword(4) Performance evaluation
Keyword(5) Uniformity of region
Keyword(6) Region boundary
1st Author's Name Masaki MATSUURA
1st Author's Affiliation Graduate School of Science and Engineering, Shizuoka University()
2nd Author's Name Keiko TAKAHASHI
2nd Author's Affiliation Graduate School of Science and Engineering, Shizuoka University
3rd Author's Name Takahiro SUGIYAMA
3rd Author's Affiliation Faculty of Information, Shizuoka University
4th Author's Name Keiichi ABE
4th Author's Affiliation Graduate School of Science and Engineering, Shizuoka University:Faculty of Information, Shizuoka University
Date 2001/5/11
Paper # PRMU2001-15
Volume (vol) vol.101
Number (no) 58
Page pp.pp.-
#Pages 8
Date of Issue