Change Point Detection

In the present paper, we develop a new efficient approach to detection of abrupt changes in a sequence of recognized digital images, which can be applied to a wide variety of practical problems of digital image processing. The principal idea of this approach consists in transforming an original sample of the observed data to a set of onedimensional statistics and measuring a distance from homogeneity. Essentially, there are two problems associated with change point detection: detecting the change and making inferences about the change point. For solving these problems, a non-parametric technique is proposed. The test for testing the null hypothesis of “no change” against the alternative of “change” is based on a version of the Waerden statistic. Estimating the change point is based on a version of the Mann-Whitney statistic. The suggested approach represents a sort of “retrospective” approach when (at each stage) one looks at a fixed sample of the statistical data and attempts to determine whether and where a change has occurred in this fixed sample of the foregoing data. The main advantage of the proposed approach lies in its relative simplicity and the ease with which it can be applied.