Please use this identifier to cite or link to this item:
https://hdl.handle.net/1959.11/15713
Title: | Fluency Function Approximation of Thin Line Images | Contributor(s): | Kawazoe, Fumio (author); Toraichi, Kazuo (author); Nakamura, Koji (author); Kwan, Paul H (author) | Publication Date: | 2003 | Handle Link: | https://hdl.handle.net/1959.11/15713 | Publication Type: | Journal Article | Source of Publication: | Gazo Denshi Gakkaishi, 32(4), p. 438-445 | Publisher: | Gazo Denshi Gakkai | Place of Publication: | Japan | ISSN: | 1348-0316 0285-9831 |
Fields of Research (FoR) 2008: | 080108 Neural, Evolutionary and Fuzzy Computation 080106 Image Processing 080109 Pattern Recognition and Data Mining |
Socio-Economic Objective (SEO) 2008: | 890202 Application Tools and System Utilities | Peer Reviewed: | Yes | HERDC Category Description: | C1 Refereed Article in a Scholarly Journal | English Abstract: | The authors have succeeded in a series of research on function approximation of raster images and their subsequent reconstruction in a scalable manner. However, these earlier research in which the target of approximation being the image contour, suffered from the quality problem of uneven line width. This problem is particularly apparent when it was applied to images having a large number of thin lines such as maps and circuit diagrams. In this paper, in order to function approximate thin line images with high quality, a novel contour tracking method is proposed that considers both characteristics of connectivity and continuity between contour segments. Furthermore, function approximation of the tracked pixel sequences by the suitable Fluency functions that include straight line, arc and second-degree curve is performed. To verify its effectiveness, the proposed contour tracking method is applied to blank map images for quality evaluation. |
---|---|
Appears in Collections: | Journal Article |
Files in This Item:
File | Description | Size | Format |
---|
Items in Research UNE are protected by copyright, with all rights reserved, unless otherwise indicated.