...
  • 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>International Journal of Innovative Computing Information and Control >VISION-BASED HYPOTHESIS INTEGRATION FOR INNER AND OUTER LIP CONTOUR DETECTION
【24h】

VISION-BASED HYPOTHESIS INTEGRATION FOR INNER AND OUTER LIP CONTOUR DETECTION

机译：基于视觉的内部和外部唇部轮廓线假说集成

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Vision-based lip contour detection is a challenging problem because lip and skin colors are similar, and the boundary between the lip and skin is usually ambiguous. We propose a real-time lip contour extraction algorithm by integrating several simple classifiers. Because the visual properties (concave-convex, shadow, illumination, and surface normal) of various parts of the lip contour vary considerably, we divided the whole lip contour into four parts (outer-upper, outer-lower, inner-upper, and inner-lower) to capture the specific characteristics of each part. The color/edge features and spatial-temporal consistency were exploited to make several simple hypotheses of lip contour pixels. For each lip contour part, a strong classifier was built by combining a set of hypotheses based on the AdaBoost algorithm to distinguish the lip contour from non-contour pixels. A deformable lip shape model was applied for fitting the lip contour by searching model parameters that maximize the classification scores along the contour. We compared the proposed algorithm with the Active Contour Model and Active Shape Model. The experiments show that both inner and outer lip contours can be detected and traced efficiently and reliably. The proposed lip contour extraction algorithm has potential for use in several fields, such as speech recognition and language learning.
机译：基于视觉的嘴唇轮廓检测是一个具有挑战性的问题，因为嘴唇和皮肤的颜色相似，并且嘴唇和皮肤之间的边界通常不明确。通过整合几个简单的分类器，我们提出了一种实时的嘴唇轮廓提取算法。由于嘴唇轮廓各部分的视觉属性（凹凸，阴影，照明和表面法线）变化很大，因此我们将整个嘴唇轮廓分为四个部分（上-上，外-下，内-上和内部-下部）捕获每个部分的具体特征。利用颜色/边缘特征和时空一致性来做出嘴唇轮廓像素的几个简单假设。对于每个嘴唇轮廓部分，通过结合一组基于AdaBoost算法的假设来构建强分类器，以区分嘴唇轮廓和非轮廓像素。通过搜索使沿轮廓的分类分数最大化的模型参数，将可变形的唇形模型应用于拟合唇形轮廓。我们将提出的算法与活动轮廓模型和活动形状模型进行了比较。实验表明，可以有效，可靠地检测和跟踪内部和外部嘴唇轮廓。提出的嘴唇轮廓提取算法具有在语音识别和语言学习等多个领域中使用的潜力。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号