影像處理技術應用於人臉膚質特徵之研究

Research on the Features of Human Skin Appearance by Image Processing

黃欣慧(Hsin-Hui Huang)

指導教授 ： 李錫捷

共同指導教授 ： 郭文嘉

元智大學/資訊學院/資訊管理學系/碩士(2006年)

https://doi.org/10.6838/YZU.2006.00197

摘要

近年來膚質檢測成為一項新興的潮流，這些非侵入性（non-invasive）之皮膚測試儀可以判斷受測者的膚齡及膚質，是一個能客觀評價人體皮膚當前狀態的工具，但皮膚測試儀每一次所能紀錄的測值僅限於探測棒所及之皮膚區塊，無法完整紀錄全臉的肌膚狀況，在實際操作之前也需要接受特別的訓練，因此若能以影像記錄皮膚狀況，並以影像處理技術找出皮膚膚質特徵，不僅能簡化操作流程，也可以針對大範圍的皮膚區塊作評估。因此本研究將以數位影像紀錄臉部肌膚，並利用影像處理技術找出皮膚表面的紋理特徵，以達到完整紀錄之目的。本研究首先會先設立一標準拍攝環境以取得原始影像，接下來會利用手動擷取我們感興趣的區域（Region of Interest, ROI），取得ROI後會分別針對基本色彩空間特性、斑點特徵、皺紋特徵、紋理特徵等四大點進行皮膚影像特徵值之運算，之後我們會利用統計學中的Pearson係數來決定哪些影像特徵值與皮膚測試儀測值、非專家觀察值、專業皮膚科醫師觀察值的關聯度最高，最後便得到與皮膚膚質測試相關度最高之特徵值。

關鍵字

影像處理 ； 膚質特徵 ； 色彩空間 ； 索貝爾邊緣偵測 ； 灰階值相互關係矩陣法 ； 長度變動編碼法

並列摘要

Skin test has been used popularly as a sort of trend recently, by which it evaluates the participant’s skin with a non-invasive way and provides information in terms of his or her condition or age of skin. Skin test instrument, however, can only record a specific range depending how far region of skin the probe can reach. Also, an operator has to be trained well before using it. Hence, skin test instrument does not have competence of providing entire information of the participant’s skin condition.
Using image processing techniques to analyze characteristics of skin and make a recording of them is not only simplifying the process of skin test, but also providing a way to widen the ability of evaluating the quality of skin. In the research, we use image processing techniques to gain distinctiveness of skin texture, and record them by a digital-imagery way. This provides the details of skin information. First at all, we set up a standard environment for photographing in order to gain the participant’s skin image as clear as possible. Then, by the region of interest (ROI) which has been drawn manually the operation will be made with the color space feature, the spot feature, the wrinkle feature, and skin texture feature to offer the features of the participant’s skin image. After this, using the Pearson statistic apprises how relevant the image feature is with the skin test instrument data, the non-expert data, and the expert data.
Conclusively, we will have the image features which are highly fit for skin evaluation.

並列關鍵字

image processing ； skin Features ； color space ； sobel operator ； gray-level co-occurrence matrix method ； run length method

參考文獻

[1] D. Chappard, A. Chennebault, M. Moreau, E. Legrand, M. Audran, and M. F. Basle, “Texture Analysis of X-Ray Radiographs Is a More Reliable Descriptor of Bone Loss Than Mineral Content in a Rat Model of Localized Disuse Induced by the Clostridium botulinum Toxin,” Bone Vol. 28, No. 1 Jan. 2001.

[2] M. Chantler, “Why illuminant direction is fundamental to texture analysis,” IEEE Proc. Vision Image Signal Process. 142, 199-206, 1995.

[3] O. G. Cula and K. J. Dana, “Skin Texture Modeling,” International Journal of Computer Vision, vol. 62, pp.97, 2004.

[4] R. W. Conners and C. A. Harlow, “Theoretical comparison of texture algorithms,” IEEE Trans. Pattern Anal. Machine Intell., vol. 2, pp.204, May 1980.

[5] S.E. El-Khamy, O.A. Abdel-Alim, and M.M. Saii, "Neural network face recognition using statistical feature and skin texture parameters," IEEE Proc. of the Eighteenth National Publication, Vol. 1, p.233-p.240, 2001.