返回Google圖書搜尋

Measuring External Face Appearance for Face Classification

出版	INTECH Open Access Publisher, 2007
ISBN	39026130339783902613035
URL	http://books.google.com.hk/books?id=2KPzoAEACAAJ&hl=&source=gbs_api

註釋In this chapter we introduce the importance of the external features in face classification problems, and propose a methodology to extract the external features obtaining an aligned feature set. The extracted features can be used as input to any standard pattern recognition classifier, as the classic feature extraction approaches dealing with internal face regions in the literature. The resulting scheme follows a top-down segmentation approach to deal with the diversity inherent to the external regions of facial images. The proposed technique is validated using two publicly available face databases in different face classification problems: gender recognition, face recognition and subject verification. In a first approach, we show that the external features encoded in the NMF coefficients yield enough useful information for classification purposes. Then we compare the information contributed by the external features and the internal features. Finally, the last step is to combine the information provided by the external and the internal features. We show that both kinds of information are complementary, providing and extra information cue that can improve the classification results in presence of occlusions and local changes in the illumination.