摘要
讨论了基于马尔可夫随机场(MRF)模型的融合颜色和边缘信息的嘴唇特征提取方法。首先进行嘴唇区域检测,再结合嘴唇形状特点建立了基于MRF的嘴唇图像分割模型,构造相应的能量函数,并采用改进的最高置信度优先(HCF)算法求解能量函数的最优解,得到图像标记场,进而提取出嘴唇轮廓。结合人脸结构信息,提出了融合鼻孔角度信息的嘴唇特征点提取方法。实验结果表明,此算法具有良好的鲁棒性。
This paper introduced a MRF（ Markov Random Field）-based method of integrating color and spatial edge information to address the problem of lip feature extraction. First, it detected a lip region using color information. Focused on this region,it constructed corresponding energy function and adopted the modified Highest Confidence First （HCF） algorithm to minimize the energy function. Then obtained the lip contour. Combining with the information of nostril,it presented a lip feature extraction method to accurately locate six key lip feature points, Experimental results demonstrate that the method is effective and robust.
出处
《计算机应用研究》
CSCD
北大核心
2007年第7期300-302,共3页
Application Research of Computers
基金
国家"973"计划资助项目(2005CCA04400)
