在所有臉部辨識與臉部偵測的研究中，提升辨識正確率與增快辨識速度是共同的研究目標。為了達到此目標，輸入影像與前處理步驟就顯得非常的重要，好的輸入影像與前處理可大大地提升後續辨識的準確率。在本文中利用臉部特徵來判斷臉部旋轉角度，再利用此計算出該角度臉部的模樣，以供辨識系統做判斷。藉由膚色區域切割與橢圓偵測，即可找出臉部位置，再依據擷取的臉部影像，利用特徵垂直的集中性，判斷人臉轉動角度及方向。而模擬計算方法為取得臉部各特徵點之數值，再利用內插法，計算出三維的虛擬臉部模型。完成上述方法的旋轉偵測，再配合虛擬資料庫的比對，即可以有限的判別資訊達到最好的判別率與最大的判別角度。

In all face recognition and face detection research, to enhance recognition accuracy with the increased speed is a rapid identification of common research goals. In order to achieve this goal, entering the image and pre-processing steps are very important. A good input image quality and pre-processing can greatly enhance the follow-up
identification accuracy. In this paper, facial features are used to determine the rotation angle of the face. And using this angle to calculate the angle of the face is in order to make judgments for the recognition system. Face location can be identified with color region cuttings and oval face detections. Captured facial images, based on the use of features of vertical concentration, determine the angle and direction of rotation face. The simulation calculation method is developed for obtaining the face value of each feature point, and calculating three-dimensional model of a virtual face by volume spline interpolation. After detecting the rotation and adding the virtual database, we can achieve the best classification rate and maximum angle in the limited information.