本篇論文提出了一個實現基於賈伯濾波器(Gabor filter)以及支持向量機(support vector machine, 簡稱SVM)的人臉識別演算法。在所提出的方法中，使用環型對稱賈伯濾波器(circularly symmetrical Gabor Filter, 簡稱CSG filter)來取代傳統的賈伯濾波器並結合SVM分類器，應用在影像識別上。 基於傳統賈伯濾波器的人臉辨識系統，對於增進人臉影像變化的穩健性是具有良好的效能，然而卻還是存在著一些問題。在傳統的賈伯濾波器中，不存在旋轉的不變性(rotation invariant)以及其計算複雜度過高，而這些問題在應用於動態及時影像識別上，會成為很大的缺點。因此，本篇論文提出使用環型對稱賈伯轉換濾波器取代傳統的賈伯轉換並且以支持向量機對影像做訓練，測試系統的排外性、濾波器參數對辨識效率的影響，以達到增進人臉識別系統的效能。 ;In this thesis, the face recognition algorithm for the realization of Gabor filter and the support vector machine (SVM) classifier is proposed. In the proposed algorithm, the circularly symmetrical Gabor Filter (SCG Filter) is used to replace the traditional Gabor filter and the SVM classifier is integrated in the application of image recognition. While the face recognition algorithm based on the Gabor filter has good performance in improving robustness on face image variances, there are still some problems. The traditional Gabor filter has no rotation invariant and its calculation is too complicated. These problems will become an obstacle in application of image recognition. In this thesis, a CSG Filter is proposed to replace the traditional Gabor transform filter and the SVM is used to train images in face recognition. Also, we test the exclusivity of the system and the impact of filter parameters on identification efficiency to promote the effectiveness of the face recognition system.
