Real-Time Face Detection Using Integral Histogram of Multi-scale Local Binary Patterns
Real-Time Face Detection Using Integral Histogram of Multi-scale Local Binary Patterns
复制标题
DOI：
10.1007/978-3-642-24728-6_37
复制
发表时间：
2011-08
期刊：
影响因子：
--
通讯作者：
Sébastien Paris;H. Glotin;Zhong-Qiu Zhao
中科院分区：
文献类型：
--
作者：
Sébastien Paris;H. Glotin;Zhong-Qiu Zhao
提出了一种新的基于多尺度局部二值模式积分直方图的快速有效的人脸检测方法。LBP及其变体的直方图是视觉系统中最先进的描述符之一，与众所周知的Bag-of-SIFT相比非常有竞争力。HMSLBP具有主要线性可分的特性，并且可以用快速大规模支持向量机（SVM）求解器有效地训练。新提出的检测器提供了一个帧速率上级到20 fps的VGA模式与现代计算机。
This paper introduces a novel fast and efficient method for face detection using integral histogram of multi-scale local binary patterns (HMSLBP). Histogram of LBP and its variants are among the state-of-the-art descriptors in vision systems and is very competitive compared to well known Bag-of-SIFT. HMSLBP has the property to be mainly linearly separable and can be trained efficiently with a fast large-scale support vector machine (SVM) solver. The new proposed detector offers a framerate superior to 20 fps in VGA mode with modern computers.