A combined analog-digital technique for normalizing video signals for the detection of moving objects

Some applications of machine vision, such as surveillance and traffic monitoring, require a system that operates without human adjustment under changing scene conditions. Such a system should adaptively condition the analog video signal to maximize the detection of interesting events. An information measure based on the probability of detection is introduced, and an adaptive scheme to manipulate the analog signal to optimize the information measure and maximize the likelihood of detection is presented.<>