標題: 結合多層次解析度之貼圖為主的特徵追蹤演算法
Multi-resolution Texture-based Feature Tracking in Large Time-varying Volume Visualization
作者: 傅光暐
Kuang-Wei Fu
莊榮宏
王才沛
Jung-Hong Chuang
Tsai-Pei Wang
多媒體工程研究所
關鍵字: 多層次解析度;特徵追蹤;時變容積;容積資料顯像;multi-resolution;feature tracking;time-varying volume data;volume rendering
公開日期: 2008
摘要: 我們提出了一個擁有多層次解析度架構 (multi-resolution hierarchy) 之有效率的貼圖特徵萃取及追蹤演算法 (feature extraction and tracking algorithm)。 在前處理運算時,我們套用蓋式濾波器 (Gabor filter) 來計算多層次架構下每層解析度的貼圖特徵係數 (texture attributes)。 各層次之特徵係數可以以其在空間上的關係,架構成一個貼圖取樣層次架構。 在執行運算時,我們使用新的基於層次架構的特徵追蹤演算法。 由於層次架構中,上下層之間有著空間上的關係,當我們從貼圖取樣架構之頂端追蹤至底層時,上層之追蹤結果,能夠幫助下層之節點決定其追蹤範圍。 在我們的結果中可以顯示多層次架構的方法明顯的改善傳統貼圖為主的特徵追蹤演算法之效率,並且此種架構也能夠良好的適應不同類型的資料及多目標追蹤 (multi-target tracking)。
We proposed an efficient texture-based feature extraction and tracking algorithm with multiresolution hierarchy. In the preprocessing step, texture attributes are computed by Gabor filtering at each level of multi-resolution hierarchy. With spatial coherence, these attributes are builded as a texture sample hierarchy. In the run-time, a hierarchical feature tacking algorithm is applied. From root to leaves of the texture sample hierarchy, we use the parents tracking result to help children decide their tracking window. We demonstrate how this structure can perform more efficient tracking in texture-based datasets. The tracking algorithm is also adapted with multi-target feature.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT009557508
http://hdl.handle.net/11536/39660
Appears in Collections:Thesis


Files in This Item:

  1. 750802.pdf