Output details

Generalized face super-resolution.

<23>This paper formulates a framework for synthesizing high-resolution face images from very low-resolution CCTV input subject to severe motion blur and changing pose and expression. Despite fictional depiction (e.g. 2006 film Déjà Vu, BBC series Spooks) this remains an open challenge. Paper is final definitive account of stream including output from EPSRC/MOD JGS grant GR/S63687/01(JGS934), and ICCV'05 (10.1109/ICCV.2005.155), CVPR'06 (10.1109/CVPR.2006.196), ICDP'05 (10.1049/ic:20050075, Best Paper Award), synthesising and extending prior work. Results include invited presentations to the UK and US governments, QinetiQ, Lockheed Martin, NICTA; development contract from UK MOD (Harbinder Rana - HSRANA@mail.dstl.gov.uk) through spin-out (rpkoger@visionsemantics.com); UK and International patents.