相片中人臉區域偵測研討與實作

機構典藏 > College of Engineering > Graduate Institute & Department of Computer Science and Information Engineering > Thesis > Item 987654321/77454

Please use this identifier to cite or link to this item: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/77454

Title:	相片中人臉區域偵測研討與實作
Other Titles:	A study and implementation of human face detection in photos
Authors:	詹智翔;Chan, Chih-Hsiang
Contributors:	淡江大學資訊工程學系碩士在職專班汪柏
Keywords:	人臉區域偵測;人臉特色;三階段偵測法;自動白平衡;形態學;連通標記;侵蝕;膨脹;Face detection;Face Feature;Tri-Stage Detection Method;Auto White Balance;Morphology;Connected-Component Labeling;Erosion;Dilation
Date:	2012
Issue Date:	2012-06-21 06:45:52 (UTC+8)
Abstract:	本論文主要是探討相片中人臉區域偵測相關技術與演算法，並且開發出一套可實現與應用的系統。首先透過自動白平衡演算法進行影像色差的調整，然後經過色彩空間轉換與膚色偵測法，萃取影像中膚色分佈的範圍，再透過影像二值化方法，可得出類似人臉之區域，為使人臉區域形狀更加完整，我們運用形態學（Morphology）裏的侵蝕（Erosion）與膨脹（Dilation）等方法，可有效消除影像中的雜訊與連通區域的擴展，再以連通標記（Connected Component Labeling）演算法，來進行膚色區域之標，最後結合人臉的形狀、臉部內容特色(Face Feature)等多種偵測方法，判定最為可能之人臉區域。經實驗結果，本研究提出的三階段偵測法(Tri-Stage Detection Method)，確實能夠精準定位出相片中複雜背景裡的人臉區域，另外加上人臉形狀偵測的結果，可準確得到人臉輪廓特色下巴的位置，進而排除非人臉區域的脖子部位，最後將修正後的人臉區域做定位與輸出，供後續之應用與處理。 This thesis is to explore the human face region detection photos related technology and algorithms, and develop a set of systems and applications can be realized. First conducted through the automatic white balance algorithms for image color adjustment, then through the color space conversion and color detection, extraction image color distribution of the range,then through the image binarization method, can be drawn from the face of similar area, to make the shape more plus a complete face region, we use morphological (Morphology) in the erosion (Erosion) and expansion (Dilation) and other methods, which can effectively eliminate the noise in the image area and connected through the expansion, then connectivity markers (Connected Component Labeling) algorithm, to carry out the skin area of the calibration, the final combination of face shape, face particular character (Face Feature) and other etection methods,to determine the most likely of the face region. The experimental results, the study proposed three-stage detection method (Tri-Stage Detection Method), is indeed able to accurately locate the photo in the complex background of the face region, plus face shape detection results may be accurate face contour characteristics chin position, ruling out non-face area of the neck area, and finally revised positioning of the face region to do with the output for the follow-up of the application and processing.
Appears in Collections:	[Graduate Institute & Department of Computer Science and Information Engineering] Thesis

Files in This Item:

File	Size	Format
index.html	0Kb	HTML	204	View/Open

Loading...