| Summary: | 碩士 === 元智大學 === 電機與資訊工程研究所 === 87 === Locating a face and recognition are two main steps in a human face recognition system. Automatic face location and recognition from a complex scene is a very difficult problem due to the highly variability which includes individual appearance, 3D pose, facial expression, and lighting. The goal of this thesis is to propose an automatic approach that can extract a face from a complex background image, and find the facial expression involving 3D pose to increase the recognition rate. The main tasks of the approach are as follows. First, each input image is preprocessed with contrast adjustment and edge detection. Secondly, wavelet decomposition transform is applied to making image into different levels. The face location is evaluated on the lowest level of the wavelet decomposition, whereas the face features appearing in the other levels are located by active contour model. This method performs the face database to gather statistics first, and then catches face features by active contour model. The varying range of an active contour is dominated by the main varieties stored in the database. On the recognition of an input face, the face's shape will be located, the related face features will be computed and compared with those in the database to identify the most match person.
|
|---|
