具適應性與快速人臉偵測方法之研究

A Study on Adaptive and Fast Face Detection

廖振墀(Jhen-chih Liao)

指導教授 ： 林文暉

崑山科技大學/資訊科技學院/資訊管理研究所/碩士(2007年)

摘要

本篇論文提出一種臉部偵測的方法，它能夠在具有不同大小人臉的一張影像中迅速且精準的偵測出臉部位置。此方法主要利用膚色像素特徵以及K-均值聚類融合（K-Means Clustering Ensembles）進行膚色分群。此方式包含三個步驟。首先是膚色像素特徵向量它包含了像素位置以及每個像素所含的色彩資訊。其次，為了提供快速且穩定的像素分類和產生高品質的區域框架，我們使用一個由一組具不同初始中心的K-均值分群（K-Means Clustering）分類器所組成的K-均值聚類融合分類器。藉由投票選舉的機制獲得膚色像素快速且穩定的分群。再考慮臉部區域所具備的幾何特性，可藉由框架整合（Frame Integration）和框架切割（Frame Segmentation）二種演算法來獲得臉部最佳化的區域範圍。這二種方法分別使用在同一臉部被分成兩個框架時進行臉部框架的結合和當不同的臉部被框在同一個區域時進行臉部框架的分割。同時去除當高度和寬度的比例超過2.3的框架即可找到一個臉部候選區域。最後，臉部確認工作可由主成分分析（Principal Component Analysis）、支援向量機（Support Vector Machine）等方法來解決。

關鍵字

臉部偵測 ； K-均值聚類融合 ； 膚色

並列摘要

In this paper a fast and robust method that is invariant to the changes of the size of the face existing in a test image. The method applies the skin color pixels features and the k-means clustering ensembles approach to cluster the skin color pixels extracted from a test image. There are three stages that are included in the proposed method. The first, the skin-color pixel feature vector included both its position and color information is extracted. For providing fast and stable skin-color pixels classification and generating high quality region frames, a k-means clustering ensembles approach which combine the clustering results obtained from a set of k-means clusters with random cluster center initialization is employed. The fast and stable partitions of skin color pixels can be obtained based on voting mechanism. Then by taking face region geometrical property into account, the optimum region boundaries are then obtained by frame integration and frame segmentation algorithms. The two algorithms are used for merging two frames that are inherently same face into a frame and partitioning a frame that has two different faces explicitly into different faces respectively. Finally, candidate face regions will be found by rejecting the framed regions when its ratio of height to width is over than 2.3. In the face verification, the human face can be verified by the principle component analysis (PCA) and support vector machine (SCM).

Keyword: face detection, k-means clustering ensembles, skin color, skin color detection.

並列關鍵字

k-means clustering ensembles ； skin color detection. ； face detection ； skin color

參考文獻

[8]S. K. Singh, D. S. Chauhan, M. Vatsa, R. Singh, “A Robust Skin Color Based Face Detection Algorithm,” Tamkang Journal, Science and Engineering, vol. 6, no. 4, pp. 227-234, 2003.

[2]Yang, J., A. Waibel, “A Real-Time Face Tracker”, Proceedings of the 3rd IEEE Workshop on Applications of Computer Vision , Vol. 26, No. 11, pp. 1408-1423, 2004.

[3]Turk, M., A. Pentland, ”Eigenface for Recognition”, Journal Cognitive Neuroscience, 3, pp. 71-86, 1991.

[4]Sirovich, L., M.Kirby, “Low-Dimensional Procedure for the Characterization of Human Faces”, Journal Optical Society America, 4, pp. 519-524, 1987.

[9]R. Byrd and RanjaniBalaji, “Real Time 2-D Face Detection Using Color Rations and K-Mean Clustering,” Database systems and computer vision, pp. 644-648, 2006.