評価関数の差分化による視覚的曲線の高速検出法

カテゴリ: 論文誌(論文単位)

グループ名: 【C】電子・情報・システム部門

発行日: 2013/05/01

タイトル(英語): Fast Detection Method for Perceptive Curves by Deformation of Evaluation Function into Recurrence Relation

著者名: 高城　貴光（岐阜大学工学部応用情報学科），斉藤　文彦（岐阜大学工学部応用情報学科）

著者名(英語): Takamitsu Takagi (Department of Information Science, Faculty of Engineering, Gifu University), Fumihiko Saitoh (Department of Information Science, Faculty of Engineering, Gifu University)

キーワード: 視覚的曲線，曲線検出，視覚的群化，ドロネー三角形分割，動的計画法　　Perceptive curve，Curve detection，Perceptual grouping，Delaunay triangulation，Dynamic programming

要約(英語): Human visual sensation has an ability finding a continuous curve from a noisy binary image. In the previous paper, we have proposed a high-precision method to detect perceptive curves from a binary image. Pre-reported method generates a graph that is accumulated evaluation values of appropriate partial curves. Then a perceptive curve is detected by global edge tracing from the graph. This method has a problem that partial curve searching requires long times. To solve this problem, an evaluation function employed in this method is deformed into a recurrence relation. Then we exhibit that Dynamic Programming can be applied to a partial curve searching algorithm. Moreover by ordering a searching-priority in the searching, calculation costs can be reduced without an additional parameter and an accuracy deterioration. Experimental results show that a processing speed of proposed method is 165 times faster than pre-reported method as the average. It is also shown that proposed method is faster than conventional methods.

本誌: 電気学会論文誌C（電子・情報・システム部門誌） Vol.133 No.5 （2013） 特集：新たな産業への応用が進む無線通信技術

本誌掲載ページ: 1047-1058 p

原稿種別: 論文／日本語

電子版へのリンク: https://www.jstage.jst.go.jp/article/ieejeiss/133/5/133_1047/_article/-char/ja/