CAJ | 학술논문

眼睛作为人脸识别最重要的生物特征之一，具有高稳定性、不易复制性和可区分性等特点，其特征提取是人脸图像检索的关键环节，通常对眼睛特征的描述主要依据视觉感知，面对由于模糊描述而无法实现量化处理的难题，提出了一种基于肤色的人眼轮廓自动提取新方法。该方法分为四步：首先，提取人脸彩色图像并将其从RGB色彩空间变换到YCbCr色彩空间；其次，通过肤色非线性空间映射进行人脸区域定位处理；然后，采用大津法实现人眼区域分割；最后，通过色度信息以及形态学处理方法提取眼睛轮廓。实验表明：该方法能够精确提取人眼轮廓，提取率为99%以上，提取时间小于1.75 s。
안정작위인검식별최중요적생물특정지일，구유고은정성、불역복제성화가구분성등특점，기특정제취시인검도상검색적관건배절，통상대안정특정적묘술주요의거시각감지，면대유우모호묘술이무법실현양화처리적난제，제출료일충기우부색적인안륜곽자동제취신방법。해방법분위사보：수선，제취인검채색도상병장기종RGB색채공간변환도YCbCr색채공간；기차，통과부색비선성공간영사진행인검구역정위처리；연후，채용대진법실현인안구역분할；최후，통과색도신식이급형태학처리방법제취안정륜곽。실험표명：해방법능구정학제취인안륜곽，제취솔위99%이상，제취시간소우1.75 s。
The eyes as a most important biometric feature of face recognition has the characteristics of high stability and dis-tinguishability,and is not easy to be copied. Its feature extraction is a key step in face image retrieval. The usual description of eye features is mainly based on visual perception. As for the defficulty that quantization processing can not be realized due to fuzzy description,a new method for eye contour automatic extraction based on skin color is put forward. The method consists of four steps:extraction of the color face image and transformation of the RGB color space into YCbCy color space,localization processing of the face region by color nonlinear space mapping,segmentation of the eye region with Otsu method,and extraction of the eye contour according to the color information and morphological processing method. The experimental results show that the method can accurately extract eye contour,the extraction rate is more than 99%,and the extraction time is less than 1.75 s.