CAJ | 학술논문

对人脸特征定位来说ASM是一个功能强大的工具.然而用作建立局部灰度结构模型时对光照非常敏感,不够鲁棒.实际上,在搜索过程中不同光照条件下训练形状模型时明显会得到不同结果.为了提高对光照变化的鲁棒性,在引入LBP算子的基础上,提出了三种基于ELBP的人脸特征定位方法:基于轮廓的ELBP-ASM,基于正方形的ELBP-ASM及基于分割正方形的ELBP-ASM.其中基于分割正方形的ELBP-ASM取得更加精确的结果,对光照变化最鲁棒.
대인검특정정위래설ASM시일개공능강대적공구.연이용작건립국부회도결구모형시대광조비상민감,불구로봉.실제상,재수색과정중불동광조조건하훈련형상모형시명현회득도불동결과.위료제고대광조변화적로봉성,재인입LBP산자적기출상,제출료삼충기우ELBP적인검특정정위방법:기우륜곽적ELBP-ASM,기우정방형적ELBP-ASM급기우분할정방형적ELBP-ASM.기중기우분할정방형적ELBP-ASM취득경가정학적결과,대광조변화최로봉.
As to positioning of facial features, ASM is a powerful tool. When used for building the model of partial gray structure, it is very sensitive to light change, and not robust enough. In fact, when training shape model, under the different light conditions, we will get different results in the searching progress. In order to improve the robustness to light change, on the basis of introducing ELBP operator, the paper puts forward three methods of facial feature's positioning based on ELBP, that is, ELBP-ASM based on profile, square and divided square separately, among which, the ELBP-ASM based on divided square can achieve more accurate result and most robustness to light change than other methods do.