Virtual Edge Extraction Method Based on New Separability

Fumihiko MORI  Hiromitsu YAMADA  Makoto MIZUNO  Naotoshi SUGANO 

Publication
D - Abstracts of IEICE TRANSACTIONS on Information and Systems (Japanese Edition)  Vol.J94-D  No.12  pp.2105-2113
Publication Date: 2011/12/01
Online ISSN: 1881-0225
Print ISSN: 1880-4535
Type of Manuscript: PAPER
Category: 
Keyword: 
edge,  separability,  segmentation of block,  statistics,  virtual edge, 

Full Text(in Japanese): PDF(15.7MB)

Summary: 
A simple robust statistical edge detection method is proposed. The proposed method is developed by combining the good points of Kundu's method where the region segmentation of local area is used and Fukui's method where a statistical evaluation value named “separability” is used for edge extraction. We developed a simple new region segmentation method of local area and a new statistical evaluation value for edge extraction and extracted a set of edge features named “virtual edge” which consists of separability, estimated edge location, edge orientation, distance between centers of segmented sub-regions and the difference of mean colors in the sub-regions. The proposed algorithm is successfully applied to many natural images and the results show that some weak points of Kundu', Fukui' and Canny' methods are improved.