A practical end-to-end test of image quality applied to digital CCTV systems with unknown compression engines

摘要

A test target pack was created by Home Office Centre for Applied Science and Technology (CAST) to aid non-technical CCTV users install a system that can be used for the observation category of 'identify'. This introduced a new concept of digital video quality testing based upon a set of computer generated facial images with measured levels of similarity and difference. After receiving feedback from various CCTV users, CAST chose to refine certain features of the test and create a second version. This report describes the development of the test target pack and also describes the method used to create and choose the final faces. After creating new faces and using subjective testing to eliminate any unsuitable faces, 12 faces were chosen as the final set: three skin tone groups with four faces in each group. The object was to create a technology agnostic test regime that would identify whether or not a system was above a 'threshold resolution' that would produce images fit for purpose.