Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/7443
標題: 磁振造影血管邊緣偵測
Vessel Boundary Detection for Magnetic Resonance Images
作者: 王宏哲
Che, Wang Hung
關鍵字: digital image processing;數位影像處理;magnetic resonance images;vessel boundary;磁振造影;邊緣偵測
出版社: 電機工程學系
摘要: 
本論文為一利用數位影像處理技術做磁振造影血管邊緣偵測的研究,最主要為希望運用其影像處理方法的一些特性,如非侵入式、非接觸性、偵測速度快且具安全性……等,來對現今磁振造影血管邊緣偵測方面有所改善,以達到節省人力和時間、增加偵測速度、獲得有用資訊及節省成本。
本論文血管邊緣偵測主要分為兩大步驟,一為權值的計算,一為血管邊緣路徑的搜尋。偵測血管邊緣最主要的方法乃是基於最短路徑搜尋的應用,希望此應用能將造影時所產生的雜訊之影響減至最小,以避免不完全或錯誤的偵測,以獲得最佳的血管邊緣位置。權值的計算部份主要為應用均化遮罩、索貝爾運算與強化運算來降低雜訊與強化邊緣的資訊,以利下一步驟邊緣搜徑的進行。而在本論文所著重的第二部份邊緣路徑搜尋方面,主要目的為取得血管邊緣位置,且在本論文中嘗試了使用四種不同的方法來做一比較。第一種方法為一點一點式的搜尋方法,此種方法的速度最快但雜訊忍受度最低,第二種方法是為了改善第一種方法所衍生出來的,採用三點三點式的搜尋,第三種方法是將Dijkstra最短路徑法應用於邊緣搜尋上,此種方法雖對雜訊忍受度最高對但搜尋速度過於緩慢。第四種方法則為改良第三種方法的缺點,同時具有速度快與雜訊忍受度高的優點。
對於本論文實驗的結果,第一種方法最快而第三種方法最慢,其中第一、二種偵測時間與圖形大小為線性增加,第三、四種則為指數增加。第四種方法為本論文最後所採用之方法,因為它同時具有第一種方法速度快的特性與第三種方法雜訊忍受度高的優點,可使我們獲得最佳的血管邊緣位置,且能節省所需時間。

The aim of this study is the vessel boundary detection for magnetic resonance images. Detecting vessel boundary usually requires two steps: The first step is weight calculation. The purpose of this step is to give a large weight to vessel boundary while suppress everything else. The second step is to search the vessel boundary form the weights calculated in the first step. The emphasis at the thesis is on the second step. We use an algorithm based on shortest distance searching. The goal is to minimize the impacts of imaging noise and artifacts. For the first step, the boundary weights are calculated by a first-order derivative operation called Sobel's operator. For the second step, four different methods are compared in this thesis. The first is to calculate the distance form one point to its neighboring points and keep only the neighboring point with the shortest distance. The second method is similar to the first method but calculate all the neighboring points in 3 levels. The third method is Dijkstra algorithm. The fourth algorithm uses level-by-level search instead of point-by-point search to
improve the speed of the third method.
Of the four methods, the first method is the fastest and the third method is the slowest. The first method is the most susceptible to noise. The third method and fourth methods are the least susceptible to noise. Since the fourth method achieves high noise-resistance capability with only a slight speed degradation. We believe that it is a
very good vessel edge detection algorithm.
URI: http://hdl.handle.net/11455/7443
Appears in Collections:電機工程學系所

Show full item record
 

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.