Face Recognition Based On Gabor Complex Amplitude Representation

With the urgent drive of the development of the electronic information fields such as social security, economic security, military criminal anti-terrorism and human-computer interaction, the face recognition technology has become an cutting-edge technology with development potential, which has also become a hot spot in the research of the computer vision field. It has important application value to let the computer with human intelligence, memory and recognition ability, and realize the real intelligent era. However, the difficulties of face recognition technology is the field of computer vision, whose main reasons are environmental factors such as illumination, expression, pose, shelter as well as behavioral factors like shooting in the face picture acquisition process. Therefore, a good face recognition algorithm should be insensitive to these factors.Because Gabor wavelet can simulate the mammalian to do the response model of simple visual cortex cell and it has excellent spatial scale and direction selectivity, which can link the pixels in an adjacent region to reflect the change of image pixels within a local scope so as to obtain the local features of face images in different scales and different directions. In addition, after the transformation of the Gabor wavelet, the face image has strong robustness to light and the local variation. Therefore, Gabor wavelet transformation has been widely used in face recognition technology feature extraction algorithm. However, in recent years, Gabor wavelet transform based feature extraction algorithm is using amplitude information, and the use of phase information of the algorithm is relatively small, because of the robust Gabor local parts of the movement, and the phase of the sensitive. But it is verified that the phase of Gabor is robust to illumination changes, so we have a thorough study and comparison on the amplitude and phase of 2D Gabor wavelet and the main work and contributions are as follows:This thesis makes a brief overview on research background and the development history of the face recognition. The in-depth study on the 2D Gabor wavelet lies a solid foundation for the research for the algorithm. And this thesis does studies, comparison and analysis on the impact of the amplitude and phase of the two-dimensional Gabor wavelet, in face recognition, on the illumination and local displacement change.This thesis does research and analysis on the algorithm of the amplitude and phase of two-dimensional Gabor wavelet transformation, including the extraction algorithm of amplitude (LGBP), extraction algorithm of phase (HGPP, LGXP, LGPDP), combining amplitude and phase (ELGBP) as well as the extraction algorithm of amplitude and phase in 3D space (GV-LBP-TOP).This thesis makes improvements on the algorithm of the LGPDP and puts forward to the ELGPDP algorithm as well as a new extraction algorithm which combines amplitude and phase, which are getting good results.This thesis shows the face recognition system project, which contains the theoretical knowledge and practical ability during the graduate years and lies a solid foundation for the life and work in the future.