Abstract
In this work, a method for selecting a subset of frames in a video is proposed, aiming to improve face recognition success rate from the video while using less number of frames. The method's goal is to select frames according to some proposed criteria which is based on positions of the facial features on frame. Positively selected frames are preprocessed with some simple image processing and dimension reduction steps and acquired data is used in a multi-class classifier. Finally, experiments done on whole set of faces and set of faces selected and preprocessed with the proposed method are presented and comparitive results are given.
Keywords
Computer science
Artificial intelligence
Preprocessor
Facial recognition system
Computer vision
Classifier (UML)
Face (sociological concept)
Pattern recognition (psychology)
Frame (networking)
Dimensionality reduction
Feature extraction
Frame rate
Subject Areas
Face and Expression Recognition
·Computer Vision and Pattern Recognition ·Physical Sciences
Face recognition and analysis
·Computer Vision and Pattern Recognition ·Physical Sciences
Image and Video Stabilization
·Computer Vision and Pattern Recognition ·Physical Sciences
Loading…