Can we ID from CCTV? Image quality in digital CCTV and face identification performance

CCTV is used for an increasing number of purposes, and the new generation of digital systems can be tailored to serve a wide range of security requirements. However, configuration decisions are often made without considering specific task requirements, e.g. the video quality needed for reliable person identification. Our study investigated the relationship between video quality and the ability of untrained viewers to identify faces from digital CCTV images. The task required 80 participants to identify 64 faces belonging to 4 different ethnicities. Participants compared face images taken from a high quality photographs and low quality CCTV stills, which were recorded at 4 different video quality bit rates (32, 52, 72 and 92 Kbps). We found that the number of correct identifications decreased by 12 (~18%) as MPEG-4 quality decreased from 92 to 32 Kbps, and by 4 (~6%) as Wavelet video quality decreased from 92 to 32 Kbps. To achieve reliable and effective face identification, we recommend that MPEG-4 CCTV systems should be used over Wavelet, and video quality should not be lowered below 52 Kbps during video compression. We discuss the practical implications of these results for security, and contribute a contextual methodology for assessing CCTV video quality.

[1]  Felicity Gibling,et al.  Artistic enhancement in the production of photo-fit likenesses: An examination of its effectiveness in leading to suspect identification , 1994 .

[2]  D. M. Green,et al.  Signal detection theory and psychophysics , 1966 .

[3]  T. Bachmann.  Identification of spatially quantised tachistoscopic images of faces: How many pixels does it take to carry identity? , 1991 .

[4]  D. M. Monro,et al.  Video surveillance using low bandwidth, high compression systems , 1997 .

[5]  Alvin G. Goldstein,et al.  The effects of discrimination training on the recognition of white and oriental faces , 1973 .

[6]  Sugato Chakravarty,et al.  Methodology for the subjective assessment of the quality of television pictures , 1995 .

[7]  V. Bruce,et al.  Face Recognition in Poor-Quality Video: Evidence From Security Surveillance , 1999 .

[8]  V. Bruce,et al.  Matching identities of familiar and unfamiliar faces caught on CCTV images. , 2001, Journal of experimental psychology. Applied.

[9]  R. Cucchiara.  Multimedia surveillance systems , 2005, VSSN@MM.

[10]  Paul Miller,et al.  Verification of face identities from images captured on video. , 1999 .

[11]  R. Malpass,et al.  Recognition for faces of own and other race. , 1969, Journal of personality and social psychology.

[12]  Vicki Bruce,et al.  Matching the faces of robbers captured on video , 2001 .

[13]  H.G. De Meer,et al.  Utility curves: mean opinion scores considered biased , 1999, 1999 Seventh International Workshop on Quality of Service. IWQoS'99. (Cat. No.98EX354).

[14]  Philip J. Benson,et al.  When does the inner-face advantage in familiar face recognition arise and why? , 1999 .

[15]  Michael C. Bromby.  To be Taken at Face Value? Computerised Identification , 2002 .

[16]  Ramin Zabih,et al.  Bayesian multi-camera surveillance , 1999, Proceedings. 1999 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (Cat. No PR00149).

[17]  H. Ellis,et al.  Identification of Familiar and Unfamiliar Faces from Internal and External Features: Some Implications for Theories of Face Recognition , 1979, Perception.

[18]  Agneta Herlitz,et al.  Higher face recognition ability in girls: Magnified by own-sex and own-ethnicity bias , 2006, Memory.

[19]  Mubarak Shah,et al.  Tracking and Object Classification for Automated Surveillance , 2002, ECCV.

[20]  C. Lewin,et al.  Sex differences in face recognition—Women’s faces make the difference , 2002, Brain and Cognition.

[21]  Wei Tsang Ooi,et al.  Critical video quality for distributed automated video surveillance , 2005, MULTIMEDIA '05.

[22]  M. Angela Sasse,et al.  Can small be beautiful?: assessing image resolution requirements for mobile TV , 2005, MULTIMEDIA '05.