Research On Gabor Transform For Face Feature Extraction

For decades, automatic face recognition technology has been an active topic in image processing and pattern recognition. Its significance lies on both theoretical research and practical applications. Gabor transform is a kind of transformation tool. It has been widely applied in face recognition, but its system for face recognition is not mature currently and in some aspects lacks basic breakthrough. This dissertation focuses on face recognition exploiting Gabor transform. The study carried out is as the following aspects:1 , Face recognition detection is put forward, base on geometric feature algorithm. The author managed to combine the face detection based on geometric features and face feature extraction based on Gabor transformto raise face recognition rate. The face recognition and detection based on geometric features is the earliest development and the most research algorithm with matured characteristics. Applied to face detection, its accuracy can exceed 95% 。 A split-merge algorithm is applied based on geometric characteristics and obtains satisfactory experiment results.2, Face feature extraction algorithm based on Gabor transform is presented. With profound research on face recognition and Gabor filter, the author tries tooptimize the filter. 40 filters and their performances on face feature extraction are surveyed. 8 common characteristics of Gabor filter are evaluated. Compared with Fourier transform, Gabor transform is proved to be better in face feature extraction.3, A fast algorithm of Gabor Transform is proposed. It is proved by experiments to be efficient in computation and fast in speed of running.4, The choice method of sample points on face characteristic sample points is evaluated. Gabor filters are selected properly and split-merge detection method is used in face feature extraction. Through Gabor filtering and dimensional reducing processes, the characteristic points of human face are extracted. The performance is proved by experiment that the recognition rate reaches 92%. The author designed framework of face recognition system based on Gabor transformis discussed and the future trend is predicted.