1 / 22

A Comparative Evaluation of Average Face on Holistic and Local Face Recognition Approaches

A Comparative Evaluation of Average Face on Holistic and Local Face Recognition Approaches. Sanqiang Zhao, Xiaozheng Zhang, and Yongsheng Gao Pattern Recognition, 2008. ICPR 2008. 19th International Conference on 8-11 Dec. 2008 Page(s):1 - 4. 授課教師：萬書言副教授

teige
Download Presentation

A Comparative Evaluation of Average Face on Holistic and Local Face Recognition Approaches

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. A Comparative Evaluation of Average Face on Holistic and Local Face Recognition Approaches Sanqiang Zhao, Xiaozheng Zhang, and Yongsheng Gao Pattern Recognition, 2008. ICPR 2008. 19th International Conference on8-11 Dec. 2008 Page(s):1 - 4 授課教師：萬書言副教授 報告者：陳奕穎 報告日期：2009/11/23

  2. Abstract Introduction Average Face Experimental settings Experimental results Conclusions References Outline

  3. Abstract • To reveal a published paper "100% Accuracy in Automatic Face Recognition" which working mechanism. • Perform the averaging process using pose-varied synthetic images generated from 3D face database and conduct a comparative study. • Two representative methods, i.e. Eigenface and Local Binary Pattern (LBP) are employed to perform the experiments.

  4. Introduction • Recently, psychological researchers from University of Glasgow proposed an “Average Face”to improve face recognition approaches. • Pose variation is confined within approximately ±30° to make all feature points visible in the image. • Averaging the 20 images of each person can increase the recognition accuracy from 54% to 100%.

  5. Average Face I • Original face images labeled with 34 feature points. • Fixed 34-point template. • Face images after morphing.

  6. Average Face II • Texture averaging. • Shape averaging.

  7. Average Face III • Average texture. • Average shape. • Average face from morphing.

  8. Eigenface Method • Eigenface Method是利用主成分分析(principal component analysis ,PCA)擷取臉部特徵，作為臉部辨識的方法。

  9. 主成分分析(Principal Component Analysis,PCA) I

  10. 主成分分析(Principal Component Analysis,PCA) II

  11. 區域性二元化圖形 (local binary patterns，LBP) I • LBP演算法是一種與灰階度無關的紋理分析演算法具有強大的辨識力。 • 擁有不依影像旋轉而改變的特性。 • LBP histogram

  12. 區域性二元化圖形 (local binary patterns，LBP) II • 在LBP演算法裡 LBP8,1是一個基本的運算。

  13. Experimental settings I • Face database • Use the USF Human ID 3D database. • Synthetic 2D images of 50 people in the database. • For each person, the facial pose is varied by incrementing the yaw angle (left-right) within ±30° range. • This yields a total of 61×50=3050 images.

  14. Experimental settings II • Face database • All the images are normalized (in scale and rotation) and cropped to 100×100 pixels based on the positions of two eyes.

  15. Experimental settings II • Evaluation protocol • In our experiments, the face images of frontal view(i.e. yaw angle is 0°) are used as the gallery set. • For the probe set, we randomly select 20 angles that are evenlydistributed within the pose range [-30°, +30°]. • Eventually we get ten groups of probe sets of face images.

  16. Experimental resultsI • Eigenface results

  17. Experimental resultsII • LBP results

  18. Experimental resultsIII • Comparative LBP results with 32 histogram bins on the set of Group 1.

  19. Experimental resultsIV • Comparative LBP results with 10x10 sized sub regions on the set of Group 1.

  20. Conclusions I • Averaging makes subtle features diluted or vanished. The two moles marked by the white boxes are indiscernible in the average face. • Experimental results show that the average face does not necessarily improve the performance of all face recognition approaches.

  21. References • [1] W. Zhao, R. Chellappa, P.J. Phillips, and A. Rosenfeld,"Face Recognition: A Literature Survey," ACM Comput-ing Surveys, 35:399-459, 2003. • [2] M. Turk and A. Pentland, "Eigenfaces for Recognition,"Journal of Cognitive Neuroscience, 3:71-86, 1991. • [3] P.N. Belhumeur, J.P. Hespanha, and D.J. Kriegman, "Ei-genfaces vs. Fisherfaces: Recognition Using Class Spe-cific Linear Projection," TPAMI, 19:711-720, 1997. • [4] L. Wiskott, J.M. Fellous, N. Krüger, and C. von der Malsburg, "Face Recognition by Elastic Bunch Graph Matching," TPAMI, 19:775-779, 1997. • [5] T. Ahonen, A. Hadid, and M. Pietikäinen, "Face Descrip-tion with Local Binary Patterns: Application to Face Recognition," TPAMI, 28:2037-2041, 2006. • [6] P.J. Phillips, W.T. Scruggs, A.J. O'Toole, P.J. Flynn,K.W. Bowyer, C.L. Schott, and M. Sharpe, "FRVT 2006 and ICE 2006 Large-Scale Results," NISTIR 7408, 2007. • [7] R. Jenkins and A.M. Burton, "100% Accuracy in Auto- matic Face Recognition," Science, 319:435-435, 2008. • [7s]R. Jenkins and A.M. Burton, "Supporting Online Mate-rial for '100% Accuracy in Automatic Face Recognition'," • Science, 319, 2008. • [8] S. Zhao and Y. Gao, "Automated Face Pose Estimation Using Elastic Energy Models," Proceedings of ICPR,4:618-621, 2006. • [9] V. Blanz and T. Vetter, "A Morphable Model for the Synthesis of 3D Faces," SIGGRAPH'99, 187-194, 1999.

  22. Thank You !

More Related