Eigenfaces Algorithm
Eigenfaces algorithm is a well-known face recognition algorithm was developed by Sirovich and Kirby (1987) and used by Matthew Turk and Alex Pentland in face classification[1]. Eigenfaces is based on a mathmatical process called principal component analysis(PCA). It uses a set of training data which are human face images and labels. Eigenvectors is solved by PCA for the face matrix, and we will get a sub-space in which the distance of different people's face images is as large as possible. Each image in testing set is projected to the sub-space, and the training image with neareast distance will be its identity.
In my work I have to add Eigenfaces module in digiKam for face recognition enhencement. In the first week, I read the code in digiKam for face recognition. I wrote a demo project to test the efficiency of Eigenfaces in OpenCV. I use orl face database which include 40 people's face images with 10 images of each person. I use 8 images of each person as traing images, and another 2 as test images, and the result is 76 true positive, 4 false positive which is acceptable.
Modification Details
My work including 4 parts:
(1) UI modification. User can choose which algorithm to use when he wants to recognize faces, so I added algorithm option in advanced tab, which is shown in the screenshot below:

(2) Algorithm Selection. The program has to use the specified algorithm which is chosen by user which including:

(a) Create a new Eigenfaces model.
(b) Read training data from face database.
(c) Add new labeled face as trainging data.
(d) Traing by calculate eigenvectors.
(e) Recognizing a new face.

(3) Database modification. I created a new table "OpenCVEigenMat" in face database to store the mat of each labeled face. Program will insert a new mat when a new face is labeled, and it will select all mat from the table when training the model.
(4) Eigenfaces algorithm. This is core part of my work. In this part, I created the Eigenfaces model, including training, recognizing, and reading/saving data from database. A threshold can be set by experience which is used to decide a face is unknown or not.
Code details
I have commited my code in a new branch. You can read and test my work there.

I used a small face data set to test, the next several screenshots are faces before recognition and after recognition.
Before recognition:

After recognition(3 left unknown):

Images in tag "Jay" after recognition:

Images in tag "Robert" after recognition:



Reference
[1] Turk, Matthew A and Pentland, Alex P. Face recognition using eigenfaces. Computer Vision and Pattern Recognition, 1991. Proceedings {CVPR'91.}, {IEEE} Computer Society Conference on 1991