Presentation 2006-10-20
Preventing Under- and Oversegmentation by Validation of Image Segments in Scale Space
Takuto KOMAZAKI, Tomoya SAKAI, Atsushi IMIYA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) This paper proposes a validation scheme for segmentation of a greyscale image. Image segmentation by edge detection is capable of extracting shapes of objects from the image. Unsupervised segmentation methods, however, have difficulty in distinguishing valid segments associated with the objects from invalid random segments due to noise. In this paper, we focus on scale space characteristics of the valid and invalid segments. Since the edge detector, such as Canny's edge detector, contains a scale parameter in its operator, the validity of the extracted segments is dependent on scale selection. Firstly, we suggest that the Canny edge can be regarded as a type of Enomoto's edge surfaces/lines for Gaussian scale-space images. Secondly, we show that the number of invalid segments monotonously decreases with increasing scale. This property allows us to identify the valid segments from coarse to fine, and to prevent under- and oversegmentation. Hierarchical relationships among the segments in the scale space help recursive evaluation of the image segment validity.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Canny edge / deep structure / critical scale
Paper # PRMU2006-111
Date of Issue

Conference Information
Committee PRMU
Conference Date 2006/10/13(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 ENG
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Preventing Under- and Oversegmentation by Validation of Image Segments in Scale Space
Sub Title (in English)
Keyword(1) Canny edge
Keyword(2) deep structure
Keyword(3) critical scale
1st Author's Name Takuto KOMAZAKI
1st Author's Affiliation Graduate School of Science and Technology, Chiba University()
2nd Author's Name Tomoya SAKAI
2nd Author's Affiliation Institute of Media and Information Technology, Chiba University
3rd Author's Name Atsushi IMIYA
3rd Author's Affiliation Institute of Media and Information Technology, Chiba University
Date 2006-10-20
Paper # PRMU2006-111
Volume (vol) vol.106
Number (no) 301
Page pp.pp.-
#Pages 6
Date of Issue