Face Detection by Discrete Gabor Jets and Reference Graph of Fiducial Points

摘要

A novel face detection scheme is described. The facial feature extraction algorithm is based on discrete approximation of Gabor Transform, called Discrete Gabor Jets (DGJ), evaluated in fiducial face points. DGJ is computed using integral image for fast summations in arbitrary windows, and by FFT operations on short contrast signals. Contrasting is performed along radial directions while frequency analysis along angular direction. Fourier coefficients for a small number rings create a long vector which is next reduced to few LDA components. Four fiducial points are only considered: two eye corners and two nose corners. Fiducial points detection is based on faceonface classifier using distance to point dependent LDA center and threshold corresponding to equal error rate on ROC. Finally, the reference graph is used to detect the whole face. The proposed method is compared with the popular AdaBoost technique and its advantages and disadvantages axe discussed.