Improved ASM Localization Algorithm for Facial Feature

被引：36

作者：

Lv, Chongshan ^{[1
]}

Zhang, Ting ^{[1
]}

Lin, Cong ^{[1
]}

机构：

[1] Beijing Inst Technol, Sch Automat, Beijing 100081, Peoples R China

来源：

2017 10TH INTERNATIONAL CONFERENCE ON INTELLIGENT COMPUTATION TECHNOLOGY AND AUTOMATION (ICICTA 2017) | 2017年

关键词：

ASM algorithm; facial feature location; PCA; local texture mode;

D O I：

10.1109/ICICTA.2017.29

中图分类号：

TP [自动化技术、计算机技术];

学科分类号：

0812 ;

摘要：

Active shape model is widely used for facial feature localization. Regarding the traditional ASM algorithm can't describe the object shape precisely, an improved ASM algorithm is proposed. At first, we establish shape model and use PCA (Principle Component Analysis) to transform high-dimensional data to lower dimensions. Another work is to establish local texture model giving sample points with different weights, and expand the search space to match facial feature points. In contrast with the traditional ASM algorithm, the improved ASM has higher accuracy.

引用

页码：100 / 102

页数：3