Application of the Karhunen-Loeve Procedure for the Characterization of Human Faces

@article{Kirby1990ApplicationOT,
  title={Application of the Karhunen-Loeve Procedure for the Characterization of Human Faces},
  author={Michael Kirby and Lawrence Sirovich},
  journal={IEEE Trans. Pattern Anal. Mach. Intell.},
  year={1990},
  volume={12},
  pages={103-108}
}
The use of natural symmetries (mirror images) in a well-defined family of patterns (human faces) is discussed within the framework of the Karhunen-Loeve expansion. This results in an extension of the data and imposes even and odd symmetry on the eigenfunctions of the covariance matrix, without increasing the complexity of the calculation. The resulting approximation of faces projected from outside of the data set onto this optimal basis is improved on average. > 

Figures from this paper

Face Processing with Whitney Reduction Networks

This paper investigates the application of the Whitney Reduction Network (WRN) to the low-dimensional characterization of digital images human faces. Motivated by Whitney's Embedding theorem from

Symmetry, probability, and recognition in face space

TLDR
Evidence is presented that the dimension of face recognition space for human faces is dramatically lower than previous estimates, and the construction of a probability distribution in face space that produces an interesting and realistic range of (synthetic) faces is constructed.

The global dimensionality of face space

  • P. S. PenevL. Sirovich
  • Computer Science
    Proceedings Fourth IEEE International Conference on Automatic Face and Gesture Recognition (Cat. No. PR00580)
  • 2000
TLDR
It is argued that, even with large statistics, the dimensionality of the PCA subspace necessary for adequate representation of the identity information in relatively tightly cropped faces is in the 400-700 range, and it is shown that a dimensionality in the range of 200 is inadequate.

Dimensionality Reduction by Sparsification in a Local-Features Representation of Human Faces

TLDR
A sparsiication strategy in the context of LFA is studied and it is shown that it gives a substantially reduced estimate of the dimension-ality of face space, 200 and is a more accurate, non-Gaussian, parameterization of the true probability density of images of human faces.

On the Dimensionality of Face Space

TLDR
Evidence is presented that talented human observers are able to identify familiar faces that lie in a space of roughly 100 dimensions and the average observer requires between 100 and 200 dimensions, which is below most current estimates.

Recognizing Faces by Weakly Orthogonalizing against Perturbations

TLDR
This paper addresses the problem of face recognition under drastic changes of the imaging processes through which the facial images are acquired with a new method that exploits the statistical information of the variations between the face image sets being compared, in addition to the features of the faces themselves.

Weak orthogonalization of face and perturbation for recognition

  • K. NagaoM. Sohma
  • Computer Science
    Proceedings. 1998 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (Cat. No.98CB36231)
  • 1998
TLDR
Experimental results on real pictures of the frontal faces from drivers' licenses show that the new algorithm improves the recognition performance over the conventional methods, and the effectiveness of the method on image sets with changes in viewing geometry is demonstrated.

Karhunen-Loeve decomposition in the presence of symmetry. I

TLDR
For a finite group, an algorithm is derived using group representation theory to reduce the cost of determining the Karhunen-Loève basis, and this technique is demonstrated on a Lorenz-type ODE system.

Representations of human faces

Several models for parameterized face representations have been proposed in the last years. A simple coding scheme treats the image of a face as a long vector with each entry coding for the intensity

Caricatural effects in automated face perception

TLDR
The problem of gender classification and identification is addressed using a geometrical description of faces, extracted automatically from digitized pictures of frontal views of people without facial hair, to perform satisfactorily on the classification tasks and exhibit the phenomenon of caricature.
...

References

SHOWING 1-10 OF 35 REFERENCES

Low-dimensional procedure for the characterization of human faces.

  • L. SirovichM. Kirby
  • Computer Science
    Journal of the Optical Society of America. A, Optics and image science
  • 1987
TLDR
A method is presented for the representation of faces that results in the characterization of a face, to within an error bound, by a relatively low-dimensional vector.

CONNECTIONIST APPROACHES TO VISUALLY-BASED FACIAL FEATURE EXTRACTION

We examine here some properties of a connectionist autoassociative matrix for storing, in a parallel and distributed fashion, face stimuli that are coded as simple patterns of spatially varying light

Machine identification of human faces

Identification of human faces

TLDR
These studies form a foundation for continuing research on real-time man-machine interaction for computer classification and identification of multidimensional vectors specified by noisy components.

Aspects of face processing

1. Introduction.- to aspects of face processing: Ten questions in need of answers.- 2. Perceptual Processes.- Microgenesis of face perception..- Recognition memory transfer between spatial- frequency

Turbulence and the dynamics of coherent structures. II. Symmetries and transformations

l. Introduction. The accurate determindtion of coherent structures depends upon having a sufficiently large database. As we will see in this part, symmetry considerations can considerably extend the

Practical Face Recognition and Verification with Wisard

TLDR
A wide range of pattern recognition problems can be solved with this approach, they include industrial inspection, speech recognition, medical pattern recognition and artificial vision.

Probability Theory I

These notes cover the basic definitions of discrete probability theory, and then present some results including Bayes' rule, inclusion-exclusion formula, Chebyshev's inequality, and the weak law of

LIII. On lines and planes of closest fit to systems of points in space

TLDR
This paper is concerned with the construction of planes of closest fit to systems of points in space and the relationships between these planes and the planes themselves.