Abstract :
In this paper, a new method of background subtraction is presented which uses the concept of a signal estimator to construct a confidence level which is always conservative and which is never better than e−s. The new method yields stronger exclusions than the Bayesian method with a flat prior distribution.