本論文提出了一種新方法以主動外觀模型（Active Appearance Model, AAM）和主成分分析法（Principal Component Analysis, PCA）為基礎，於人臉影像上的鏡框資訊之弱化處理。首先，Adaboost將人臉影像從背景中擷取出來，使得大部分的背景資料被分離；第二步，以AAM進行人臉特徵分析，以形狀與紋理之結構關係對人臉特徵有模組化的描述，所以可以得知眼睛、嘴唇和鏡框的位置，根據AAM搜尋的結果進行人臉影像的正規化和鎖定包含鏡框區域，並且採用調適閥值方法（Adaptive Threshold）進行鏡框區塊的強化，被強化出來的鏡框區塊將被標記成損毀的影像區域；第三步，為了補償損毀的影像區域，採用PCA之影像重建，產生一張沒配戴眼鏡的人臉影像，所以損毀的影像區域會以PCA重建之影像相對應的區域進行補償，弱化原來的鏡框資訊，合成一張鏡框區塊被弱化的人臉影像；最後，採用遞迴處理重覆進行上述的影像合成動作，提升合成影像的品質，最後產生一張鏡框資訊最低的合成影像。本論文之實驗採用Eigenface之人臉辨識方法來驗證鏡框資訊被弱化之合成影像，並且拍攝粗框眼鏡和細框眼鏡的人臉影像來進行實驗比較。根據實驗的結果，本論文的方法的確能有效的提升配戴眼鏡之人臉影像在人臉辨識的辨識率，尤其在粗框眼鏡之人臉影像上，傳統的Eigenface的辨識率會受到嚴重影響，但是本論文的方法可以弱化鏡框資訊，使得合成影像的辨識率顯著提高。由於遞迴處理逐漸地弱化鏡框區塊的特徵，包含鏡框資訊、鏡框產生的陰影區域和鏡片的反光資訊，使得合成影像的品質提高，並且在人類視覺上比較自然。

In this thesis, a novel method for impairing eyeglasses features from a human facial image based on Active Appearance Model (AAM) and Principal Component Analysis (PCA) is proposed. Firstly, a human face image is captured and detected by using the Adaboost. A human face image can thereby be separated from the background image. Second, the features of a human face image are modeled using the AAM with the information of shapes and textures so that the locations of eyes, mouth and eyeglasses are known. The results of AAM searching are used to focus on the region containing eyeglasses, and then the method of Adaptive threshold is adopted to intensify the eyeglasses block. Third, a simple PCA reconstruction method is used to compensate the eyeglasses using the color of skin. As a result, a human face image without eyeglasses can be constructed. Therefore, the eyeglasses block can be replaced with the corresponding block of the constructed image and then a synthesized image impairing eyeglasses features is built. Finally, the quality of the synthesized image is further improved by recursive synthesized process. After that, an optimal synthesized human face image with only a little remaining information of eyeglasses is obtained. In this thesis, the eigenface face recognition method is chosen to verify that the synthesized image impairing eyeglasses features extracted by our proposed method can indeed enhance the face recognition performance. Besides, human face images wearing the heavy frame eyeglasses and the light frame eyeglasses are also analyzed in the experimental result. Experimental results show that our proposed method provides an effective solution to resolve the problem of eyeglasses occlusion, especially for the case of human face images wearing heavy frame eyeglasses in which the face recognition rate will decrease rapidly. The proposed method can impair the eyeglasses features and thereby improve the face recognition rate effectively. Moreover, the quality of synthesized image perceived by the human looks more natural for human vision.

[1] B. W. Hwang and S. W. Lee, “Reconstruction of Partially Damaged Face Images Based on a Morphable Face Model,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 25, pp. 365-372, March 2003.
[2] V. Blanz and T. Vetter, “A Morphable Model for the Synthesis of 3D Faces,” International Conference on Computer Graphics and Interactive Techniques, pp. 187-194, July 1999.
[3] A. M. Martínez, “Recognizing Imprecisely Localized, Partially Occluded, and Expression Variant Faces from a Single Sample per Class,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 24, pp. 748-763, June 2002.
[4] Y. Saito, Y. Kenmochi, and K. Kotani, “Estimation of Eyeglassless Facial Images Using Principal Component Analysis,” International Conference on Image Processing, Vol. 4, pp. 197-201, October 1999.
[5] M. A. Turk and A. P. Pentland, “Face Recognition Using Eigenfaces,” IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pp. 586-591, June 1991.
[6] X. Jiang, M. Binkert, B. Achermann, and H. Bunke, “Towards Detection of Glasses in Facial Images,” International Conference on Pattern Recognition, Vol. 2, pp. 1071-1073, August 1998.
[7] Z. Jing and R. Mariani, “Glasses Detection and Extraction by Deformable Contour,” International Conference on Pattern Recognition, Vol. 2, pp. 933-936, August 2000.
[8] H. Wu, G. Yoshikawa. T. Shioyama, S. Lao, and M. Kawade, “Glasses Frame Detection with 3D Hough Transform,” International Conference on Pattern Recognition, Vol. 2, pp. 346-349, August 2002.
[9] Y. Saito, Y. Kenmochi, and K. Kotani, “Extraction of a Symmetric Object for Eyeglass Face Analysis Using Active Contour Model,” International Conference on Image Processing, Vol. 2, pp. 231-234, September 2000.
[10] C. Du and G. Su, “Eyeglasses Removal from Facial Images”, Pattern Recognition Letters, Vol. 26, pp. 2215-2220, October 2005.
[11] P. Viola and M. J. Jones, “Rapid Object Detection Using a Boosted Cascade of Simple Features,” IEEE Computer Society Conference on Computer Vision and Pattern Recognition, Vol. 1, pp. 511-518, 2001.
[12] P. Viola and M. J. Jones, “Robust Real-Time Face Detection,” International Journal of Computer Vision, Vol. 57, pp. 137-154, May 2004.
[13] T. F. Cootes, G. J. Edwards, and C. J. Taylor, “Active Appearance Models,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 23, pp. 681-685, June 2001.
[14] T. F. Cootes, C. J. Taylor, D. H. Cooper, and J. Graham, “Active Shape Models—Their Training and Application,” Computer Vision and Image Understanding, Vol. 61, pp. 38-59, January 1995.
[15] G. J. Edwards, A. Lanitis, C. J. Taylor, and T. F. Cootes, “Statistical Models of Face Images—Improving Specificity,” Image and Vision Computing, Vol. 16, pp. 203-211, March 1998.
[16] G. J. Edwards, C. J. Taylor, and T. F. Cootes, “Interpreting Face Images Using Active Appearance Models,” International Conference on Face & Gesture Recognition, pp.300, April 1998.
[17] G. J. Edwards, C. J. Taylor, and T. F. Cootes, “Learning to Identify and Track Faces in Image Sequences,” International Conference on Computer Vision, pp.317, January 1998.
[18] J. S. Park, Y. H. Oh, S. C. Ahn, and S. W. Lee, “Glasses Removal from Facial Image Using Recursive Error Compensation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 27, pp. 805-811, May 2005.
[19] M. B. Stegmann, B. K. Ersbøll, and R. Larsen, “FAME - A Flexible Appearance Modelling Environment,” IEEE Transactions on Medical Imaging, Vol. 22(10), 2003.