Authors:
Kamal Nasrollahi
and
Thomas B. Moeslund
Affiliation:
Aalborg University, Denmark
Keyword(s):
Face Quality Measures, Super Resolution, Face Logs.
Related
Ontology
Subjects/Areas/Topics:
Applications
;
Computer Vision, Visualization and Computer Graphics
;
Enhancement and Restoration
;
Feature Extraction
;
Features Extraction
;
Image and Video Analysis
;
Image Formation and Preprocessing
;
Image Quality
;
Implementation of Image and Video Processing Systems
;
Informatics in Control, Automation and Robotics
;
Motion, Tracking and Stereo Vision
;
Pattern Recognition
;
Real-Time Vision
;
Signal Processing, Sensors, Systems Modeling and Control
;
Software Engineering
;
Tracking of People and Surveillance
;
Video Analysis
Abstract:
Using faces of small sizes and low qualities in surveillance videos without utilizing some super resolution algorithms for their enhancement is almost impossible. But these algorithms themselves need some kind of assumptions like having only slight motions between low resolution observations, which is not the case in real situations. Thus a very fast and reliable method based on the face quality assessment has been proposed in this paper for choosing low resolution observations for any super resolution algorithm. The proposed method has been tested using real video sequences.