查詢結果分析
來源資料
頁籤選單縮合
題名 | 基於局部區域膚色再分析之多重人臉偵測=Multiple Face Detection Based on the Reanalysis of Human Skin Color in Local Area |
---|---|
作者姓名(中文) | 蔡吉昌; 危永中; 吳佳純; 崇達明; | 書刊名 | 黃埔學報 |
卷期 | 67 2014.11[民103.11] |
頁次 | 頁77-87 |
分類號 | 312.13 |
關鍵詞 | 人臉偵測; 人臉辨識; 光線補償; 索貝爾邊緣檢測; Face detection; Face recognition; Light compensation; Sobel edge detection; |
語文 | 中文(Chinese) |
中文摘要 | 靜態影像中的人臉偵測是人臉辨識的基石,而人臉辨識是許多應用需要的技術,例如:影像資料的分類與搜尋,門禁安全的管理及罪犯的搜尋與追蹤等。然如何讓處理影像的電腦知道靜態影像中是否有人臉的出現實是一件不容易的事,尤其影像中的人臉重疊時,會破壞人臉的外型,使得人臉的外型特徵無法利用而造成辨識效果大打折扣。 本研究的目的在於研究如何正確地偵測出靜態影像中多重的人臉,包含重疊的人臉,論文中除運用了一般常用的膚色分析、光線補償、及形態學的侵蝕與膨脹等技術外,本研究並提出局部區域的膚色再分析方法,並利用無空洞區域的影像統計特性,以進一步的去除近似人臉膚色的雜訊區域,之後利用索貝爾邊緣檢測及嘴唇頻色分析來分辨相連之人臉,以實現多重人臉之偵測目的。 |
英文摘要 | Face detection applied to still images is the cornerstone of face recognition, and face recognition technologies are needed for many applications, such as: classification and search of images, security management, and search and track of criminals. However, it is not really easy to detect faces in still images for computers. Specially, when faces overlap, face detection techniques will be unable to utility characteristics of the face shape. Consequently, the effect of face detection will be greatly reduced. The purpose of this research is to study how to properly detect multiple faces in still images, including overlapping faces. In addition to some techniques commonly used by face detection, for instance skin color analysis, light compensation and morphological erosion and expansion, etc., this study proposes the methods of reanalyzing human skin color in local area and using statistical characteristics of non-empty area of the image to further remove noisy area with the approximate of the face skin color. Afterwards, Sobel edge detection and the analysis of lip color are used to distinguish overlapping faces for achieving the purposes of multiple-face detection. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。