檢測對於現代科技生活非常熱門，而這篇文章裡我們對於皮膚的檢測，是利用統計上的貝氏定理的模型(model)來檢測皮膚，分出是皮膚的部分或非皮膚的部分，這是目前非常有效而且也很準確的方法。首先我們收集許多臉部的照片，並是先區分出皮膚與非皮膚的部分，這個動作我們稱之完全認知(global knowledge)，然後將模型所有的數學符號都以相對應的物理量來解釋，並利用MATLAB程式語言，統計及計算出所有機率值，而這些數值都將儲存起來當作參考資料，並以這些資料建構貝氏濾波器。新的照片利用此模型計算後，來做比較，歸納判斷出照片中皮膚與非皮膚的部位；結果顯示判斷與真實的數值吻合，亦接近我們所期待的結果。根據此初步研究的結果，我們計畫未來將加上其他的條件來增加此濾波器的準確性。 Skin detection is a useful technology for digital image processing. In this paper, we used the Baye's filter to separate the area of skin and non-skin area. The Bayes' theorem is a powerful statistical processing. Our method has first analyzed collected photo and manually selected the skin and non-skin area of image. This process represents that we have a so called ”global knowledge” toward to the understanding of human skin or other area before detecting the target photos. We analyzed and built a statistical model for two distinct areas which their physical meanings are related to human skin or non-skin. We wrote Matlab codes to calculate the posterior probabilities of two areas and found the results are matched the real situation for our photos. The results of our detection process showed that the rate of truly identifying of skin for real skin area is high. Also our process missed some skin area. But almost none of non-skin area has been identified as skin. The primitive results showed the promise of our method, further study and expansion of the proposed method will be continued.
