Open AccessFace Tracking via Content Aware Correlation Filter
Author(s) -
Houjie Li,
Shuangshuang Yin,
Fuming Sun,
Fasheng Wang
Publication year - 2021
Publication title -
international journal of circuits, systems and signal processing
Language(s) - English
Resource type - Journals
ISSN - 1998-4464
DOI - 10.46300/9106.2021.15.76
Subject(s) - computer science , facial motion capture , artificial intelligence , computer vision , tracking (education) , histogram , face (sociological concept) , feature (linguistics) , video tracking , locality , face detection , facial recognition system , histogram of oriented gradients , pattern recognition (psychology) , feature extraction , object (grammar) , image (mathematics) , psychology , pedagogy , social science , linguistics , philosophy , sociology
Face tracking is an importance task in many computer vision based augment reality systems. Correlation filters (CFs) have been applied with great success to several computer vision problems including object detection, classification and tracking, but few CF-based methods are proposed for face tracking. As an essential research direction in computer vision, face tracking is very important in many human-computer applications. In this paper, we present a content aware CF for face tracking. In our work, face content refers to the locality sensitive histogram based foreground feature and the learning samples extracted from complex background. It means that both foreground and background information are considered in constructing the face tracker. The foreground feature is introduced into the objective function which could learn an efficient model to adapt to the face appearance variation. For evaluating the proposed face tracker, we build a dataset which contains 97 video sequences covering the 11 challenging attributes of face tracking. Extensive experiments are conducted on the dataset and the results demonstrate that the proposed face tracker shows superior performance to several state-of-the-art tracking algorithms.