眼鏡去除與人臉辨識

Eyeglasses Removal and Face Recognition

陳律臻(Lu-Chen Chen)

指導教授 ： 曾定章

國立中央大學/資訊電機學院/資訊工程學系/碩士(2012年)

摘要

人臉辨識 (face recognition) 與人臉偵測 (face detection) 技術被廣泛地應用在各種實務系統上，像是門禁系統、監控系統、身分認證的登入系統等。然而，人臉辨識的辨識率常常受到許多因素影響，包含光照環境的不同、表情的不同、臉部旋轉、及有遮蔽物的情形等。因此，我們針對帶眼鏡的議題，提出眼鏡去除的方法，希望能提高人臉辨識系統的辨識率。
眼鏡去除系統主要包含三個部分，人臉偵測、眼鏡偵測、及眼鏡去除。首先，利用膚色偵測與 Adaboost 偵測從影像中找出人臉區塊。接著，再根據眼鏡的特性，從膚色二值化影像、與邊偵測二值化影像，定位出眼鏡區域，並參考主成份分析法 (PCA) 重建影像將五官部分去除。最後，我們以 PCA 重建影像為參考影像，將其膚色部分的平均數與標準差與戴眼鏡影像膚色部分的平均數與標準差做顏色的對應調整，之後，將眼鏡部分以調整後的 PCA 重建影像中對應位置的膚色填補，再以兩張影像的差異值做細部調整，以及邊界平滑化，產生眼鏡去除後的影像。
在眼鏡去除的應用上，我們找了兩種人臉辨識方法做實驗測試，分別針對無戴眼鏡影像、戴眼鏡影像、與眼鏡去除影像進行人臉辨識率的分析。在實驗中，可以發現戴眼鏡影像是會影響辨識系統的辨識率的，而針對戴眼鏡的人臉影像，在臉部辨識前，先去除眼鏡，再辨識人臉，也是可以提高臉部辨識率，減少受到眼鏡的影響。

關鍵字

人臉偵測 ； 眼鏡偵測與眼鏡去除 ； 人臉辨識

並列摘要

Face recognition and face detection techniques are widely used in various applications, such as access control systems, surveillance system, login system, etc. However, there are some factors that affect the recognition performance like different lighting conditions, expression, size, and occlusion by other objects. In this thesis, we propose an eyeglasses removal method to generate a naturally looking glassless facial image to improve the face recognition rate.
The proposed system consists of three modules: face detection, eyeglasses detection, and eyeglasses removal. First, we detect the face regions by skin color and Adaboost detection. Second, we use skin color and edge data to detect the eyeglasses region. Finally, we generate an eyeglasses- removal facial image by PCA reconstruction method and color correction strategy.
After all, we combine the eyeglasses removal system to face recognition system, for face recognition result of the images of face without eyeglasses, with eyeglasses, and eyeglasses removal. We can find that the proposed method can improve the recognition rate for overall face recognition system.

並列關鍵字

face detection ； eyeglasses detection ； eyeglasses removal ； face recognition

參考文獻

[1] Andrew, A. M., "Another efficient algorithm for convex hulls in two dimensions," Information Processing Letters, vol.9, no.5, pp.216-219, 1979.

[2] Bay, H., A. Ess, T. Tuytelaars, and L. V. Gool, "SURF: speeded up robust features," Computer Vision and Image Understanding, vol.110, no.3, pp.346-359, 2008.

[3] Belhumeur, P. N., J. P. Hespanha, and D. J. Kriegman, "Eigenfaces vs. Fisherfaces: recognition using class specific linear projection," IEEE Trans. on Pattern Analysis and Machine Intelligence, vol.19, no.7, pp.711-720, 1997.

[4] Cootes, T. F., 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, 1995.

[5] Dreuw, P., Steingrube, P., Hanselmann, H. and Ney, H., "SURF-face: face recognition under viewpoint consistency constraints," in Proc. British Machine Vision Conf., London, UK, Sep.7-10, 2009.