Elsevier

Pattern Recognition

Volume 41, Issue 8, August 2008, Pages 2710-2717
Pattern Recognition

Computation of a face attractiveness index based on neoclassical canons, symmetry, and golden ratios

https://doi.org/10.1016/j.patcog.2007.11.022Get rights and content

Abstract

Analysis of attractiveness of faces has long been a topic of research. Literature has identified many different factors that can be related to attractiveness. In this research we analyze the role of symmetry, neoclassical canons, and golden ratio in the determination of attractiveness of a face. We focus on the geometry of a face and use actual faces for our analysis. We find there are some differences in the criteria used by males and females to determine attractiveness. The model we have developed to predict the attractiveness of a face using its geometry is accurate with low residual errors.

Introduction

A popular axiom concerning physical attractiveness is: “Beauty is in the eye of the beholder”. Research in the area of facial perception has identified many different factors that contribute to a face being considered attractive. Armstrong [1] suggests that beauty cannot be defined by one single principle. Rhodes [2] focuses on averageness, symmetry, and sexual dimorphism and their link to facial attractiveness. Little et al. [3] suggest that self-perceived attractiveness influences one's opinion of the attractiveness of others, and DeBruine [4] shows both males and females prefer faces that resemble their own.

In this paper, we develop a quantitative method for measuring facial attractiveness using a combination of several factors that have been deemed significant in previous research. Many previous studies have used composite faces or faces that are altered in some other way to study the effects of symmetry and averageness on attractiveness [2], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16]. In contrast, we use the actual faces compiled from a standard face recognition database for our analysis as the process of averaging faces for composites can diminish the appearance of attributes that would classify a face as more or less attractive [17]. We then determine the location of important landmarks in the face [18], [19]. In all, 29 landmarks on each face as described by Shi et al. [20] are used to take physical measurements and compute the values of three factors: neoclassical canons, symmetry, and golden ratios. The faces are presented to a set of human subjects to determine their perceived attractiveness to find which factor or which combination of factors is the best predictor of attractiveness. We systematically investigate the relationship between a face's measurements and its attractiveness. The features that play the greatest role in attractiveness are identified for both genders of raters and faces. In addition, the way males and females view attractiveness in faces of the same and opposite gender is explored as there are differing accounts in the literature. Perrett et al. [12] show that males and females prefer caricatured composite faces to average composite faces in both male and female images. O’Toole et al. [17] find that females rate female faces significantly higher than they rate male faces and that the femininity of females is strongly related to attractiveness. Full details of this research can be found in Ref. [21].

Section snippets

Datasets and experimental design

We begin with an image database containing a set of face images for the experiment and analysis. Using the image database, two datasets are compiled for the analysis. The feature dataset consists of the locations of the landmarks in the faces. The attractiveness dataset contains the attractiveness ratings given to the images by the human participants.

Computation of attractiveness predictors

The main motivation of this research is to examine the attractiveness of a face, Fi, as a function of its face geometry captured by a set of m landmarks. Thus: Fi={fi1,fi2,,fim},where each feature point, fij=(xij,yij), 1in, 1jm, is represented by its two-dimensional spatial coordinates in the face. The goal is to determine a function A that maps a face to an attractiveness score. A(Fi)[1,10].To compute the attractiveness, we use three predictors that have been proposed in literature:

Analyses and results

We begin with the examination of a set of general questions about the attractiveness of human faces, including the variability of raters and effect of self-perceived attractiveness. Then the roles of the three predictor variables, neoclassical canons, symmetry, and golden ratio, in the attractiveness of a face are examined. All analyses use the statistical analysis software (SAS) [33].

Summary and future work

The goal of this study is to determine a predictive model for attractiveness based on neoclassical canons, symmetry, and golden ratios. In contrast with much of the previous work, our study used landmarks and geometry based means for computing symmetry and had people rate actual faces instead of composite or altered faces. We also include both faces of the general population and known attractive faces. In addition, both the gender of the rater and the face are identified as to compare the

About the AuthorKENDRA SCHMID obtained a M.S. (2004) and Ph.D. in Statistics (2007) from the University of Nebraska-Lincoln. Besides her interests in pattern and shape analysis and statistics education, her research has involved statistical computing and nonlinear models. She is an Assistant Professor of Biostatistics at the University of Nebraska Medical Center.

References (38)

  • A.J. O’Toole et al.

    3D shape and 2D surface textures of human faces: the role of “averages” in attractiveness and age

    Image Vis. Comput.

    (1999)
  • D.I. Perrett et al.

    Symmetry and human facial attractiveness

    Evol. Hum. Behav.

    (1999)
  • J. Shi et al.

    How effective are landmarks and their geometry for face recognition?

    Comput. Vision Image Understanding

    (2006)
  • J. Armstrong

    The Secret Power of Beauty: Why Happiness is in the Eye of the Beholder

    (2004)
  • G. Rhodes

    The evolutionary psychology of facial beauty

    Annu. Rev. Psychol.

    (2006)
  • A.C. Little et al.

    Self-perceived attractiveness influences human female preferences for sexual dimorphism and symmetry in male faces

    Proc. R. Soc. Lond. Ser. B. Biol. Sci.

    (2000)
  • L.M. DeBruine

    Facial resemblance increases the attractiveness of same-sex faces more than other sex faces

    Proc. R. Soc. Lond. Ser. B. Biol. Sci.

    (2004)
  • R. Kowner

    Facial asymmetry and attractiveness judgement in developmental perspective

    J. Exp. Psychol. Hum. Percept. Perform.

    (1996)
  • J.H. Langlois et al.

    Attractive faces are only average

    Psychol. Sci.

    (1990)
  • J.H. Langlois et al.

    What is average and what is not average about attractive faces?

    Psychol. Sci.

    (1994)
  • A.C. Little et al.

    The role of masculinity and distinctiveness in judgments of human male facial attractiveness

    Br. J. Psychol.

    (2002)
  • D.I. Perrett et al.

    Effects of sexual dimorphism on facial attractiveness

    Nature

    (1998)
  • D.I. Perrett et al.

    Facial shape and judgements of female attractiveness

    Nature

    (1994)
  • G. Rhodes et al.

    Facial symmetry and the perception of beauty

    Psychol. Bull. Rev.

    (1998)
  • G. Rhodes et al.

    Are average facial configurations attractive only because of their symmetry?

    Psychol. Sci.

    (1999)
  • G. Rhodes et al.

    Averageness, exaggeration, and facial attractiveness

    Psychol. Sci.

    (1996)
  • J.P. Swaddle et al.

    Asymmetry and human facial attractiveness—symmetry may not always be beautiful

    Proc. R. Soc. Lond. Ser. B. Biol. Sci.

    (1995)
  • A.J. O’Toole et al.

    The perception of face gender: The role of stimulus structure in recognition and classification

    Mem. Cognit.

    (1997)
  • L.G. Farkas

    Anthropometry of the Head and Face

    (1994)
  • Cited by (0)

    About the AuthorKENDRA SCHMID obtained a M.S. (2004) and Ph.D. in Statistics (2007) from the University of Nebraska-Lincoln. Besides her interests in pattern and shape analysis and statistics education, her research has involved statistical computing and nonlinear models. She is an Assistant Professor of Biostatistics at the University of Nebraska Medical Center.

    About the AuthorDAVID MARX is a Professor of Statistics at the University of Nebraska-Lincoln. He obtained his B.S. in Chemistry from the College of Wooster in Ohio, M.S. and Ph.D. in Statistics from the University of Missouri and the University of Kentucky, respectively. His research interests include spatial statistics, geostatistics, and statistical computing.

    About the AuthorASHOK SAMAL received his B.Tech. in Computer Science from Indian Institute of Technology and Ph.D. from the University of Utah. He is an Associate Professor with the Department of Computer Science and Engineering at the University of Nebraska-Lincoln. He has published over 70 papers in image understanding, document analysis, and geospatial computing.

    View full text