基底細胞瘤(BCC)最常見於臺灣與新加坡[1]。預防與早期診斷對於人民很重要。電腦輔助診斷(CAD)在皮膚癌上已有許多進展與應用。CAD有節省計算時間與一致客觀結果的優點。
皮膚痣圖片毛髮辨識去除是CAD的第一步驟。本論文著重在，針對本論文的臨床圖片，優化改進一被廣泛使用的除毛軟體，DullRazor®。毛髮辨識的準確度影響去除毛髮與保留皮膚皺紋特徵之間的取捨。
DullRazor®的效能取決於圖片的大小、背景皮膚的顏色、毛髮粗細等。而本論文資料圖片的毛髮較細且圖的中心皺紋有白色反光。為了針對本論文的資料增進辨識程度，對相關參數作進一步的檢視與設計客製化的演算法是必要的。本論文提出以中值濾波器演算法來去除圖片中央光反射所造成的雜訊。為了公平地與DullRazor®的方法比較，本論文探討形態學斷開與閉合的結構元素的最佳大小與形狀。另外，相於傳統數位方向性分數，本論文亦提出一解析方向性分數作比較。
總結，中值濾波器演算法效能較DullRazor®方法適用。但是，毛髮辨識的效能變因多樣。客制化調整參數以增進辨識效能是必要的。而本論文可被作為調整參數時的參考流程。

Basal cell carcinomas (BCCs) are more common in Taiwan and Singapore [1]. The prevention and early detection are important to people’s health. Studies on Computer-aided diagnosis (CAD) have developed to assist diagnosis on skin cancers. CAD has the advantages of speed and objective outcomes.
Hair removal is the first step for computerized analysis on lesion images. This thesis aims at improving the performance of a popular body hair removal method, DullRazor®, on our data images. Accuracy of the detection is crucial for the balance between removing hair and removing wrongly other skin features.
The performance of DullRazor® depends on the dimension of images, background skin color, and the thickness of hair. The features of most images are small in length and width, and contain light reflection in the center. In order to peak the performance for the data images of this thesis, it is necessary to look into the properties of parameters and modify the procedure according to the data images.
This thesis proposes the median filter method for the hair removal of the data images with light reflection in the center. To compare with DullRazor method fairly, this thesis also customizes the size and shape of the structure element of morphological closing and opening for the data images. Lastly, a generalized analog directional score is introduced to compare with the conventional digital one for reducing noise.
In summary, the median filter method performs better than the open-close method, thus better than the DullRazor method. However, the performance of hair detection is sensitive to the properties of the images. It is suggested to adjust some parameters when higher performance is required. The procedure of this thesis could serve as an example to fine-tune the parameters systematically.

1. Moser S, Borm J, Michic-Probst D, Jacobsen C, Kruse-Gujer AL. Metastatic basal cell carcinoma: report of a case and review of the literature. Oral Surg Oral Med Oral Pathol Oral Radiol. 2013. doi:pii:S2212-4403(12)00398-7.
2. A. Huang, W.Y. Chang, H.Y. Liu, and G.S. Chen, "Capillary detection for clinical images of basal cell carcinoma," in Proc. ISBI, pp. 306-309, 2012.
3. Doi K. Computer-aided diagnosis in medical imaging: historical review, current status and future potential. Comput Med Imaging Graph. 2007;31:198-211.
4. Cancer Research UK. Skin cancer statistics: www.cancerresearchuk.org (accessed December 2010).
5. Lomas A, Leonardi-Bee J, Bath-Hextall F. A systematic review of worldwide incidence of nonmelanoma skin cancer. Br J Dermatol 2012;166:1069-80.
6. The prevention, diagnosis, referral and management of melanoma of the skin: concise guidelines (Newton Bishop J, Bataille V, Gavin A, Lens M, Marsden J, Mathews T, Ormerod A, Wheelhouse C). Royal College of Physicians and British Association of Dermatologists. Concise guidance to good practice series, No 7. London : RCP, September 2007
7. Kittler H., Pehamberger H., Wolff K., Binder M., "Diagnostic accuracy of dermoscopy," Lancet Oncol. 3(3), 159-165 (2002). doi: 10.1016/S1470-2045(02)00679-4
8. Scott Bicheno. Global Smartphone Installed Base Forecast by Operating System for 88 Countries: 2007 to 2017.
9. Meera Dalal, M.D. "Can a Smartphone Do What Your Doctor Does?" ABC News. Apr 27, 2013.
10. Lee T, Ng V, Gallagher R, Coldman A, McLean D. Dullrazor?: a software approach to hair removal from images. Computers in Biology and Medicine. 1997;27(6):533-543.
11. Kimia Kiani, Ahmad R. Sharafat, E-shaver: An improved DullRazor? for digitally removing dark and light-colored hairs in dermoscopic images, Computers in Biology and Medicine, Volume 41, Issue 3, March 2011, Pages 139-145
12. An Introduction to Morphological Image Processing by Edward R. Dougherty, ISBN 0-8194-0845-X (1992)
13. Hanley JA, McNeil BJ. The meaning and use of the area under the Receiver Operating Characteristic (ROC) curve. Radiology 1982 143 29-36
14. Slue W, Kopf AW, Rivers JK. Total-Body Photographs of Dysplastic Nevi. Arch Dermatol. 1988;124(8):1239-1243. doi:10.1001/archderm.1988.01670080051017.
15. E. Arias-Castro and D.L. Donoho, "Does median filtering truly preserve edges better than linear filtering?", Annals of Statistics, vol. 37, no. 3, pp. 1172–2009.