由粗到细的渐进式特征点定位算法  

Accuracy Increasing Lip Feature Point Location Algorithm

作　　者：贾熹滨[1] 石勤[1] 尹宝才[1] 

机构地区：[1]北京工业大学计算机学院多媒体与智能软件技术北京市重点实验室,北京100022

出　　处：《北京工业大学学报》2006年第5期447-450,共4页Journal of Beijing University of Technology

基　　金：国家自然科学基金资助项目(60375007).

摘　　要：为了解决描述口形轮廓的特征点定位问题,提出了一种特征点定位算法,通过建立口形灰度图像与口形特征点矢量之间的线性映射关系,实现了特征点的初始定位．在此基础上利用局部纹理模型,对特征点位置进行了准确调整．该算法克服了单纯基于局部纹理模型可能陷入搜索过程中局部收敛的缺点．并提出了适合描述唇部特征的局部纹理模型,提高了特征点标定的准确性．In order to solve the feature points location problem of describing the month appearance, a kind of feature location algorithm was proposed. It realized the preliminary feature point location through establishing the affine relationship between lip gray scale image and lip feature point vector. Based on it, the local texture model is used to adjust every point＇s accurate location. The feature point location algorithm the paper proposed overcomes the shortage of traditional local-texture-model-based algorithm, which may get in the local convergence. The paper proposed a kind of local texture model which is suitable to describing the lip appearance, and which improves the accuracy of feature point location.

关 键 词：图像分析 模式匹配 特征定位 奇异值分解 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]