| 摘要: | 在本論文中，我們提出的系統主要是由三個部分組成。第一個部分是人臉偵測。第二部分是人臉角度分類。第三部分是人臉辨識。第一個部分「人臉偵測」分成兩個小部分，第一個小部分主要是藉由皮膚顏色分割和等腰三角形與直角三角形為基礎來搜尋潛在臉的區域。然後，藉由人臉權值面具函數確認每一個人臉的正確位置。在本系統中，我們能夠同時處理灰階及彩色影像並成功地偵測正面人臉與側面人臉，且容許人臉有不同的尺寸及不同的光線與表情，並有約百分之九十八的成功率。第二個小部分我們使用HSI色彩空間並利用橢圓人臉模型去偵測每張候選人臉。我們提出了一個結合色彩及形狀的方法來偵測人臉位置。實驗結果亦成功地節省複雜背景之NTSC-RGB彩色影像約百分之七十的執行時間(約百分之十的執行時間花在色彩空間之轉換：NTSC-RGB 轉換為 HSI)。第二部分是「人臉角度分類」，我們首先將第一個部分「人臉偵測」之第一個小部分偵測到的人臉區域，再藉由方向權值面具函數判斷人臉的正確方向。最後，再藉由角度權值面具函數決定人臉轉的角度。實驗結果透露約有百分之九十九的成功率。第三部分是「人臉辨識」，我們首先將第一個部分「人臉偵測」之第一個小部分偵測到的人臉區域，再藉由特徵函數算出(1) 人臉眼睛部份區域的黑像素數目。(2) 人臉鼻子部份區域的黑像素數目。(3) 人臉嘴巴部份區域的黑像素數目。(4) 人臉全部區域的權值。(5) 人臉眼睛部份區域的權值。(6) 人臉鼻子部份區域的權值。(7) 人臉嘴巴部份區域的權值。最後，再藉由投票分數完成人臉辨識。 In this dissertation, the problems of face detection, pose classification, and face recognition are studied and solved completely. The applications of face detection, pose classification, and face recognition are extended to various topics. The applications include: computer vision, security system, authentication for remove banking and access-control application. In the past, the problems of face detection, pose classification, and face recognition were introduced by numerous researches. However, experimental results reveal the practicability and competence of our proposed approaches in finding human face, pose classification, and face recognition. The feasibility and efficiency of our approaches is confirmed by experimental results. In this thesis, the relationship between two eyes and one mouth is shown clearly based on the geometrical structure of an isosceles triangle. The first proposed face detection system consists of two primary parts. The first part is to search for the potential face regions. The second part is to perform face verification. This system can conquer different size, different lighting condition, varying pose and expression, and noise and defocus problems. In addition to overcome the problem of partial occlusion of mouth and sunglasses, the system can also detect faces from the side view. Experimental results demonstrate that an approximately 98 % success rate is achieved. In addition, a new method of extracting the human-skin-like colors is proposed for reduction of the total computation effort in complicated surroundings. In this approach, skin-color-segmentation is used to remove the complex backgrounds according to the values of R, G, and B directly. This partition method reveals the skin-color-segmentation, which results in the saving of the total computation effort nearly by 80% in complicated backgrounds. The third chapter presents another novel face detection algorithm that is presented to locate multiple faces in color scenery images. A binary skin color map is first obtained in the color analysis stage. Then, color regions corresponding to the facial and non-facial areas in the color map are separated with a clustering-based splitting algorithm. Thereafter, an elliptic face model is devised to crop the real human faces through the shape location procedure. Last, local thresholding technique and a statistic-based verification procedure are utilized to confirm the human faces. The proposed detection algorithm combines both the color and shape properties of faces. In this work, the color span of human face can be expanded as wilder as possible to cover different faces by using the clustering-based splitting algorithm. Experimental results also reveal the feasibility of our proposed approach in solving face detection problem. The fourth chapter presents a method for automatic estimation of the poses/degrees of human faces. The proposed system consists of two primary parts. The first part is to search the potential face regions that are gotten from the isosceles-triangle criteria based on the rules of "the combination of two eyes and one mouth". The second part of the proposed system is to perform the task of pose verification by utilizing face weighting mask function, direction weighting mask function, and pose weighting mask function. The proposed face poses/degrees classification system can determine the poses of multiple faces. Experimental results demonstrate that an approximately 99 % success rate is achieved and the relative false estimation rate is very low. The fifth chapter presented a robust and efficient feature-based classification to recognize human faces embedded in photographs. The proposed system consists of two main parts. The first part is to detect the face regions. The second part of the proposed system is to perform the face recognition task. The proposed face recognition system can handle different size and different brightness conditions problems. Experimental results demonstrate that we can succeed overcome the various brightness conditions. Finally, conclusions and future works are given in Chapter 6. |
