跳到主要內容

臺灣博碩士論文加值系統

(18.205.192.201) 您好!臺灣時間:2021/08/05 04:12
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

我願授權國圖
: 
twitterline
研究生:蔡博藝
研究生(外文):Tsai, Bor Yih
論文名稱:骨幹理論之整合與新方法的設計
論文名稱(外文):A Unified Theory of Image Skeleton and Its Application on Method Design
指導教授:周本生
指導教授(外文):Chow, Ben Shen
學位類別:碩士
校院名稱:國立中山大學
系所名稱:電機工程研究所
學門:工程學門
學類:電資工程學類
論文種類:學術論文
論文出版年:1994
畢業學年度:82
語文別:英文
論文頁數:49
中文關鍵詞:圖訊識別影像骨幹自動化撿測細化
外文關鍵詞:Image skeletonPattern recognitionAuto-inspectionThinning
相關次數:
  • 被引用被引用:0
  • 點閱點閱:116
  • 評分評分:
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:0
由於骨幹對圖形有提綱挈領的作用,所以在影像處理的運用領域如自動化
檢測﹑醫學圖診斷中,骨幹化是一關鍵技術。目前骨幹的開發皆以快速為
目標,而以形似為滿足,準確性卻無法涉及,且骨幹的定義有各種版本,
甚至連細化也被廣泛地誤認為骨幹.使骨幹方法間更難作效益的比較,而
應用的目標也變得糢糊,在方法設計時不能準確掌握。本論文將對於骨幹
的理論做通盤之研究與整合,首先我們將骨幹的三種版本定義證明其互等
,三種版本以觀念簡略來分各別是火燒(fire pro- pagation),大圓(
maxdisk)與內切圓(inscribed disk)在證明過程中,三種定義的特性得以
闡釋。火燒的概念生動但不易掌握;內切圓定義明確但不利推展作法;大
圓定義艱澀但據此可推出數理形態學 (mathematic mo- rphology )的方
法。證明完成主要是藉著我們新提出的輔助定義─即第四種版本,〞局
部(local )〞最遠的定義,這個定義可直接推出距離轉換 (distance
transform)求取骨幹之方法,更重要的是由〝局部(local)〞的觀念使得
我們能夠解決目前骨幹法則的問題:現行法則找到的骨幹是斷裂不完整的
。骨幹理論的整合,使定義與方法得以連接,各種不同方法的骨幹因而有
了比較的統一標準,且局部最遠的定義提供了新的設計的方向,我們並據
此設計出一個新方法。
Skeletons are the briefest forms for image patterns. Thus ske-
letonization is a key technique for pattern recognition
applied to fields such as auto-inspection and medical image
diagnosis. However, there exist some confused concepts for
digital skeleton . This is due to the fact of existance of
different forms of de- finitions for skeleton and so many look-
alikes, which are usual- ly regarded as skeletons, for
instance, results of thinning alg- orithm. Moreover the
skeletons obtained by current algorithms are disconnected.
This is because they do not use the correct definition for
skeleton,i.e., fire front''s definition. In this thesis, the
three fundamental definitions, classified to be based upon the
concepts of fire propagation, maxdisk and inscribed disk
respectively, are unified to a concepts of dista- nce to the
boundary and are thereby proved to be equivalent. In the
process of proof, the use of the definition and its concept
behind is stressed and become clear. The proof is
accomplished by the introduction of a new definition which is
based upon the concept of local maximum for distance to
boundary. By this defi- nition, the method by distance
transform can be deduced directly . Further more, the accuracy
of skeleton can be achieved by sel- ecting the local
neighborhood to meet the requirement of the se- lected distance
measure function. On the other hand, we will also analyze
and compare the performances of current algorithms for
skeletonization. The analysis of the algorithm is aimed on
paralleling their strategies of achieving a same goal in
order to compare their differences in implementation concepts.
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top