HMM Face Recognition Algorithm using Garbor filter and Hidden Markov Model

摘要

PURPOSE: A face recognition algorithm using a Gabor filter and an HMM(Hidden Markov Model) is provided to correspond to a variation of an object flexibly by manufacturing a filter independent of an illumination variation using a Gabor Wavelet, performing a discreteness of data, and using an HMM which is a probability method. CONSTITUTION: An object of an image on a necessary frequency is extracted using a Gabor Wavelet. Cosine and sine values are fixed as pi/2 and calculated, and a boundary surface of the image corresponded to a horizontal component is extracted. A binary-coded is performed by applying a threshold in the extracted image. A candidate area is selected based on a distributed area. Only face area of the image selected on the candidate area is collected and an inherent value of a matrix is obtained using an Eigenfaces method. A multiply calculation by an Eigenfaces is performed by subtracting each area candidate from an average image. If the calculated value has reliability, the area becomes a face area.