影像處理研究領域裡，人臉辨識這個議題受到廣泛的重視。在本論文中，建構一套使用局部特徵的相似人臉影像檢索系統，本系統可透過網際網路上傳人像圖片進行檢索搜尋，系統架構主要可分人臉偵測、特徵擷取及人臉影像檢索三大部分，人臉偵測使用Viola與Jones[1]提出了積分影像(Integral Image)概念和一個基於Adaboost 方法訓練人臉偵測分類器的方法。特徵擷取部分則將人臉各特徵區域預先劃分比例，再透過Haar-like特徵的基礎上，經由Boosting演算法學習訓練(左眼、右眼、鼻子及嘴巴)參數，再與特徵區域進行Adaboost的特徵偵測，得到的特徵利用主成分分析(PCA)方法抽取特徵參數。最後人臉影像檢索則為選擇檢索特徵部位及權重值設定所組成，透過複合式的特徵部位檢索資料庫得到相似人臉影像。 In the field of image processing research, the topic of face recognition has been catch the extensive attention. In this paper, we use local features to build a Similar Face Image Retrieval System. This system can search portraits by uploading it to the Internet. The system architecture can be mainly divided into three parts, including face detection, feature capture and facial characteristic searching. The face detection part takes advantages of the Integral Image concept from Viola [1] and a human face detecting training method of Adaboost. The feature capture part pre-scales each facial feature, and adds parameters through Boosting algorithm (left eye, right eye, nose and mouth) on the base of Haar-like characteristics. Then we combine the first and second part to detect the feature capture regions by Adaboost. The detected features were analyzed by principal component analysis (PCA) method to extraction of characteristic parameters. Finally, the Face Image Retrieval System is consisting of alternate feature regions and the value set weight. By using this system, we could search similar face images through a feature database.
