Video Surveillance: Legally Blind?

This paper shows that most surveillance cameras fall well short of providing sufficient image quality, in both spatial resolution and colour reproduction, for the reliable identification of faces. In addition, the low resolution of surveillance images means that when compression is applied the MPEG/JPEG DCT block size can be such that the spatial frequencies most important for face recognition are corrupted. Making things even worse, the compression process heavily quantizes colour information disrupting the use of pigmentation information to recognize faces. Indeed, the term `security camera' is probably misplaced. Many surveillance cameras are legally blind, or nearly so. Keywords-CCTV; surveillance; face recognition; image quality; spatial resolution; compression;

• Type: Conference
• Year: 2009
• Where: DICTA
• Authors: Peter Kovesi
• Keywords: DICTA 2007, DICTA 2009, Face Recognition, Spatial Resolution, Surveillance Cameras