摘要
在本篇論文中，我們提出了一個新的偵測人臉方法。我們所提出的方法包括兩個階段:訓練階段與測試階段。在訓練階段，我們使用Sobel的測邊運算、型態學的運算、以及閥值擷取將一張影像形成邊的影像(edge image)。然後，利用MAFIA演算法去探勘這些邊的影像或非邊的影像以得到這些訓練影像的最大頻繁樣式(maximal frequent pattern)，並產生正向特徵樣式(positive feature pattern)與負向特徵樣式(negative feature pattern)。在測試階段，我們利用滑動視窗法在測試影像的任何位置偵測不同大小的人臉。對於每個視窗，我們計算此視窗之邊的影像與特徵樣式的Hausdorff距離。如果此距離小於預先設定的閥值，我們接著檢查此一視窗是否包含正向特徵樣式中絕大多數的部份。如果一個視窗通過以上所有檢查就被認為是人臉。實驗結果顯示出，我們的方法在MIT-CMU的測試資料庫與我們自己的測試資料庫中分別達到98.35％與95.45％的偵測率，勝過由Schneiderman與Kanade所提出的方法。
並列摘要
In this thesis, we propose a novel face detection method based on the MAFIA algorithm. Our proposed method consists of two phases. In the training phase, we first apply Sobel’s edge detection operator, morphological operator, and thresholding to each training image, and transform it into an edge image. Then, we use the MAFIA algorithm to mine the maximal frequent patterns from those edge images and obtain the positive feature pattern. Similarly, we can obtain the negative feature pattern from the non-edge images, each of which is a complement of an edge-image. In the detection phase, we apply a sliding window to the test image in different scales. For each sliding window, we first compute the modified Hausdorff distances between the edge image of the sliding window and the feature patterns obtained. If the distances are less than the predefined thresholds, we check if the edge image of the sliding window contains most components of the positive feature pattern. If yes, the sliding window is considered as a human face. The experimental results show that our method achieves 98.35% detection rate in the MIT-CMU database and 95.45% in our own database, and outperforms the method proposed by Schneiderman and Kanade.