1、PDF外文：http:/ 1 大连民族学院毕业设计 外文资料翻译 所 在 学 院： 通信与 信息工程学院 专 业 (班级 )： 通信 工程（ 072） 学 生 姓 名： 曹小南 指 导 教 师： 许 爽 &nbs
2、p; 2011 年 5 月 9 日 2 A New Method in Locating and Segmenting Palmprint into Region-of-Interest C. Poon, D.C.M. Wong, H.C. Shen Dept. of Computer Science The Hong Kong University of Science & Technology Hong Kong helenscs.ust.hk Abstract Various tech
3、niques in analyzing palmprint have been proposed but to the best of our knowledge, none has been studied on the selection and division of the region-ofinterest (ROI). Previous methods were always applied only to a fixed size square region chosen as the central part of the palm, which were then divid
4、ed into square blocks for extraction of local features. In this paper, we proposed a new method in locating and segmenting the ROI for palmprint analysis, where the selected region varies with the size of the palm. Instead of square blocks,the region is divided into sectors of elliptical half-rings,
5、 which are less affected by misalignment due to rotational error. More importantly, our arrangement of the feature vectors ensures that only features extracted from the same spatial region of two aligned palms will be compared with each other. Encouraging results obtained favor the use of this metho
6、d in the future development of palmprint analysis techniques. 1. Introduction Biometrics, identification of a person by his/her physiological or behavioral characteristics, has become increasingly prevalent in modern identification and verification systems 12. Of all the biometrics studied, palmprin
7、t has an advantage over other biometrics such as voice and face recognition where uniqueness between people is doubtful 3 or fingerprint and iris pattern where high-resolution images are required (e.g. over 400 dpi). Palmprints are unique between people and relatively lowresolution images will suffi
8、ce (less than 100 dpi) 45. However, there is a major shortfall in the previous algorithms. That is they only utilize a fixed area of a palm for identification regardless of the actual palm size. Obviously, a lot of information has been overlooked. Two palms are captured from the same distance and wi
9、th the same resolution. A fixed size region-of-interest (ROI) (red dotted square) is considered to be small on large palms. 3 Figure 1. Palm size difference in two individuals In this paper, we propose a new method in palmpri
10、nt identification that allows the ROI in analyzing a palm image to vary with the actual palm size (see Figure 1, blue solid square). In the following sections, related work in the study of palmprint is presented in Section 2. Section 3 provides the details of our proposed method. Section 4 discusses
11、 the experimental setup with the results shown in Section 5. Section 6 concludes the major findings and contributions of this paper. 2. Related Work The studies of palmprints were first carried out on inked palmprint images 6. Not only that the image collection process is tedious and unrealistic, th
12、e hollowed central part of the palm is often missing. With the advance of technology, inkless palmprint images can now be captured, by either scanning technologies or CCD cameras. Palm images captured by CCD cameras, which is the technology used in this study, are of lower resolution but in turn req
13、uired less processing time. Apart from utilizing the structural properties of palmprint 67, other typical features that have been studied include: fuzzy directional element energy features, adopted from Chinese characters recognition algorithms8； features in Fourier space 9； eigen features retained
14、after performing dimensionality reduction by Karhunen-Loeve transform 10 or Fishers linear discriminant 11； and, statistical features obtained by texture analysis using Gabor filters 4 or wavelets transform 512. In all these previous studies, subjects are required to place their hands on a contact surface. Some even fixed subjects hand posture by pegs 4. Features are only extracted from a fixed size square block of the central part of the palm. Since
