Detection of Small Objects in Noisy Images with Unknown Probability Distributions

Abstract

The problem of detecting small-size objects in noisy images in the absence of information on the probability distributions of the magnitudes of the signals observed at points in the region of the object and at points of the complex (random) background surrounding the object is considered. Appropriate critical values are obtained for the relatively recently introduced new non-parametric statistical test in the case of detection of a small-size object. The powers of the new test and the well-known two-sample Wilcoxon test are compared for the case with exponential distributions.