Journal of the Korean Institute of Intelligent Systems (한국지능시스템학회논문지)

Korean Institute of Intelligent Systems (한국지능시스템학회)
권호 표지
DOI QR코드

DOI QR Code

An Efficient Facial Expression Recognition by Measuring Histogram Distance Based on Preprocessing

전처리 기반 히스토그램 거리측정에 의한 효율적인 표정인식

  • Cho, Yong-Hyun (School of Computer and Information Comm. Eng., Catholic Univ. of Daegu)
  • 조용현 (대구가톨릭대학교 공과대학 컴퓨터정보통신공학부)
  • Received : 2009.07.20
  • Accepted : 2009.09.30
  • Published : 2009.10.25
Download PDF
⟨ Previous Next ⟩

Abstract

This paper presents an efficient facial expression recognition method by measuring the histogram distance based on preprocessing. The preprocessing that uses both centroid shift and histogram equalization is applied to improve the recognition performance, The distance measurement is also applied to estimate the similarity between the facial expressions. The centroid shift based on the first moment balance technique is applied not only to obtain the robust recognition with respect to position or size variations but also to reduce the distance measurement load by excluding the background in the recognition. Histogram equalization is used for robustly recognizing the poor contrast of the images due to light intensity. The proposed method has been applied for recognizing 72 facial expression images(4 persons * 18 scenes) of 320*243 pixels. Three distances such as city-block, Euclidean, and ordinal are used as a similarity measure between histograms. The experimental results show that the proposed method has superior recognition performances compared with the method without preprocessing. The ordinal distance shows superior recognition performances over city-block and Euclidean distances, respectively.

본 논문에서는 전처리 기반 히스토그램 거리측정에 의한 효율적인 얼굴표정 인식기법을 제안하였다. 여기서 전처리는 중심이동과 히스토그램 평활화에 의해 인식성능을 개선하기 위함이고, 히스토그램 사이의 거리측정은 영상 상호간의 유사도를 측정하기 위함이다. 특히 중심이동은 1차 모멘트 평형에 기반을 둔 것으로 불필요한 배경을 제거시켜 위치나 크기 변화에 강건한 인식을 위함뿐만 아니라 거리의 측정부하를 줄이기 위함이다. 히스토그램 평활화는 조명의 세기에 의한 영상의 명암대비 감소에 강건한 인식을 위함이다. 제안된 기법을 320*243 픽셀의 72개(4명*18장) 표정얼굴을 대상으로 히스토그램 사이의 유사도 측정을 위해서 city-block, Euclidean, 그리고 ordinal 거리를 각각 이용하였다. 실험결과, 제안된 기법은 중심이동 및 히스토그램 평활화의 전처리를 거치지 않는 기법보다 우수한 인식성능이 있으며, ordinal 거리가 가장 높은 인식성능이 있음을 확인하였다.

Keywords

References

  1. S. H. Cha and S. N. Srihari, 'On Measuring the Distance between Histogram,' Pattern Recognition, Vol. 35, pp. 1355-1370, 2002 https://doi.org/10.1016/S0031-3203(01)00118-2
  2. F. Serratosa and A. Sanfeliu, 'Signatures versus Histograms : Definitions, Distances and Algorithms,' Pattern Recognition, Vol. 39, pp. 921-934, 2006 https://doi.org/10.1016/j.patcog.2005.12.005
  3. 조용현, 디지털 영상처리 실무, 도서출판인터비젼, 2005년 2월
  4. F. D. Jou, K. C. Fan, and Y. L. Chang, 'Efficient Matching of Large-size Hstograms,' Pattern Recognition, Vol. 25, pp. 277-286, 2004 https://doi.org/10.1016/j.patrec.2003.10.005
  5. T. Kailath, 'The Divergence and Bhattacharyya Distance Measures in Signal Selection,' IEEE Trans., Comm. Technology, COM-15, No. 1, pp. 52-60, 1967
  6. K. Matusita, 'Decision Rules, Based on the Distance for Problems of Fit, Two Samples and Estimation,' Ann. Math. Statistics, Vol. 26, pp. 631-640, 1955 https://doi.org/10.1214/aoms/1177728422
  7. S. H. Cha, 'Taxonomy of Nominal Type Histogram Distance Measures,' American Conference on Applied Mathematics, Harvard, Massachusetts, USA, pp. 325-330, Mar. 2008
  8. M. H. Yang, D. Kriegman, and N. Ahuja, 'Detecting Faces in Images: A Survey,' IEEE. Trans. on Pattern Analysis and Machine Intelligence, Vol. 24, No. 1, pp. 64-58, Jan. 2002
  9. S. H. Jeng, H. Y. M. Liao, C. C. Han, M. Y. Chern, and Y. T. Liu, 'Facial Feature Detection Using Geometrical Face Model: An Efficient Approach', Pattern Recognition, Vol. 31, No. 3, pp. 273-282, 1998 https://doi.org/10.1016/S0031-3203(97)00048-4
  10. Peter Eisert and Bernd Girod, 'Analyzing Facial Expressions for Virtual Conference,' IEEE Computer Graphics and Applications, Vol. 18, No. 5, pp. 70-78, Sept. 1998 https://doi.org/10.1109/38.708562
  11. 김상철역, 재료역학, 청문출판사, 1992년 4월
  12. 'Yale Face Databases,' http://cvc.yale.edu/projects/yalefaces/yalefaces.html