Hierarchical Tree Structured Model for Facial Landmark Detection and Tracking

• Main Authors: Shih-Chieh Huang, 黃士傑
• Other Authors: Gee-Sern Hsu
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/80896548414860263292

碩士 === 國立臺灣科技大學 === 機械工程系 === 103 === Although the Tree Structured Model (TSM) is proven effective for solving face detection, pose estimation and landmark localization in an unified model, its sluggish runtime makes it unfavorable in practical applications, especially when dealing with cases of multiple faces. We propose the Hierarchical Tree Structured Model (HTSM) to improve the run-time speed and localization accuracy. The HTSM is composed of two component TSMs, the coarse TSM (c-TSM) and the refined TSM (r-TSM), and a Bilateral Support Vector Regressor (BSVR). The c-TSM is built on the low-resolution octaves of samples so that it provides coarse but fast face detection. The r-TSM is built on the mid-resolution octaves so that it can locate the landmarks on the face candidates given by the c-TSM and improve precision. The r-TSM based landmarks are used in the forward BSVR as references to locate the dense set of landmarks, which are then used in the backward BSVR to relocate the landmarks with large localization errors. The forward and backward regression goes on iteratively until convergence. We also researched in different feature extractions, including HOG, LBP and DCT, then we give a conclusion of performance and speed about the three features. The performance of the HTSM is validated on three benchmark databases, the Multi-PIE, LFPW and AFW, and compared with the latest TSM to demonstrate its efficacy.