The eigenfacerecognition approach was developed by Turk and Pentland (1991), both colleagues from MIT, in 1987.
The eigenfeatures system measures the distance between these points on a live face and compares them to the sets of eigenfeatures stored in the database to determine whether the face is a match (Randall, 1999).
The eigenface approach reduces the amount of data needed to identify an individual to 1/1000th of a full sized image (Lau Technologies, 1999).
Eigenface - encyclopedia article about Eigenface. Free access, no registration needed. What does Eigenface mean? What ...(Site not responding. Last check: 2007-10-20)
Eigenfaces are a set of eigenvectors In linear algebra, the eigenvectors (from the German eigen meaning "inherent, characteristic") of a linear operator are non-zero vectors which, when operated on by the operator, result in a scalar multiple of themselves.
Probably because of the importance of its role in social interaction, psychological processes involved in face perception are known to be present from birth, complex, involve large and widely distributed areas in the brain and can be selectively damaged to cause a specific impairment in understanding faces known as prosopagnosia.
Some subsequent eigenfaces can be seen to correspond to generalized features such as left-right and top-bottom asymmetry, or the presence or lack of a beard.
Each eigenface represents only certain features of the face, which may or may not be present in the original image.
If, contrary, the particular feature is not (or almost not) present in the original image, then the corresponding eigenface should contribute a smaller (or not at all) part to the sum of eigenfaces.
Eigenfaces with low eigenvalues can be omitted, as they explain only a small part of characteristic features of the faces.
It is centered around the creation of the "eigenface" basis for "face space." It also discusses simplifying the eigenface basis to a level that is both managable and accurate.
The eigenfaceface recognition system can be divided into two main segments: creation of the eigenface basis and recognition, or detection, of a new face.
The eigenfaces that we are looking for are simply the eigenvectors of C. However, since C is of dimension N (the number of pixels in our images), solving for the eigenfaces gets ugly very quickly.
Recognition is performed by projecting a new image into the subspace spanned by the eigenfaces (“face space”) and then classifying the face by comparing its position in face space with the positions of known individuals.
Eigenfaces are ghostly looking face images that researchers describe as the most efficient way to encode the face, bearing an uncanny resemblance to the criminal composites of the famous nineteenth-century eugenicist Frances Galton.
Image features used in the eigenface technique need not correspond to our intuitive notions of facial features, the MIT researchers explain, but the process does ostensibly resemble human face perception in that recognition occurs quickly using a representation of the whole face.
Since composites are produced from the original database images and inherit all their feature locations from them, the composites maintain the eye alignment and general structure of the originals.
The Eigenface method (as applied to the mug shot search problem) is based on the presumption that the correlation between the Eigenface and human metrics for determining distance (or similarity) is a strong one.
Since we know there is some correlation between the Eigenface similarity metric and the human one [HBB97], we might guess that the closest image in Eigenface space (of the 100) would regularly show up somewhere among the user's top five database choices.
I implemented the eigenfaces algorithm for face recognition as outlined in class and in the Turk and Pentland paper.
This is likely due to the face that subject 11 is very unique within the data set, largely due to her hair and skin tone, while subject 01 is much closer to an 'average' subject.
As you can see from these results, only the bronze statue of Elvis was incorrectly classified as a face, possibly because it consists of a darkish blob over a generally lighter background, like many of the face images.
While undoubtedly successful in some circumstances, the theoretical foundation for the use of eigenfaces is less clear.
It is thus natural to seek a more principled justification for the use of eigenface, or at least subspace codings.
The use of larger scale versions of certain face-features in addition to the whole face as an input to a ``normal'' eigenface-based system can also be considered as another way of approximating the shape-free transformation.
Eigenface or Principal Component Analysis (PCA) methods have demonstrated their success in face recognition, detection, and tracking.
The representation in PCA is based on the second order statistics of the image set, and does not address higher order statistical dependencies such as the relationships among three or more pixels.
The representation in the Eigenface approaches is based on the second order statistics of the image set, i.e., covariance matrix, and does not use high order statistical dependencies such as the relationships among three or more pixels.
Face recognition software typically works by using a set of eigenfaces, which are essentially standardized facial features derived from a statistical analysis of many pictures of faces.
The principal eigenface in a set looks like a fuzzy averaged androgenous human face.
Early work on computerized face recognition was done in the 1960's; the first work on eigenfaces was done in 1989.
------- dot product used to determine degree of correlation or similarity between eigenface Uk and the input image after the average face is subtracted That is, wk is a number that describes how similar the image X is to the ith eigenface feature image.
In general, if there are enough eigenface images, then we can use them to reconstruct exactly the original image as follows: X = (summation from i=1 to m of wi*Ui + A) Hence, the image X is a linear combination of the eigenface images using the wi's as the weights.
If entire pattern tree is verified that corresponds to one hypothesized match position at root, then output the corresponding pixel coordinates in the level 0 image, indicating that is the position where the face was found.
