Statistical model-based change detection in moving video

Abstract

in: Signal Processing. See also BIBT E X entry below. BIBT E X:@article{AAC93a, author = {Til Aach and Andr\'e Kaup and Rudolf Mester}, title = {Statistical Model-Based Change Detection in Moving Video}, journal = {Signal Processing}, publisher = {Elsevier}, volume = {31}, number = {2}, year = {1993}, pages = {165--180}}This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by the authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author's copyright. These works may not be reposted without the explicit permission of the copyright holder. Abstract. A major issue with change detection in video sequences is to guarantee robust detection results in the presence of noise. In this contribution, we first compare different test statistics in this respect. The distributions of these statistics for the null hypothesis are given, so that significance tests can be carried out. An objective comparison between the different statistics can thus be based on identical false alarm rates. However, it will also be pointed out that the global thresholding methods resulting from the significance approach exhibit certain weaknesses. Their shortcomings can be overcome by the Markov random field based refining method derived in the second part ofthis paper. This method serves three purposes: it accurately locates boundaries between changed and unchanged areas, it brings to bear a regularizing effect on these boundaries in order to smooth them, and it eliminates small regions if the original data permits this.Zusammenfassung. Ein wichtiger punkt bei der Anderungsdetektion in Videosequenzen ist die Robustheit der Detektionsresultate gegenüber Rauschen. In diesem Beitrag werden zuerst verschiedene Teststatistiken in dieser Hinsicht miteinander verglichen. Die Verteilungen dieser Teststatistiken bei gegebener Nullhypothese werden angegeben, so daß Signifikanztests durchgeführt werden können. Ein objektiver Vergleich der verschiedenen Teststatistiken kann dann auf der Basis gleicher Fehldetektionsraten vorgenommen werden. Es werden aber auch einige Unzulänglichkeiten der aus dem Signifikanzansatz resultierenden globalen Schwellwertmethode aufgezeigt. Diese Schwächen können durch die im zweiten Teil beschriebene, auf Markov-Zufallsfeldern basierende Methode zur Verfeinerung von Anderungsmasken ausgeglichen werden. Diese Methode verfolgt drei Ziele: die Grenzen zwischen geänderten und ungeänderten Regionen werden genau lokalisiert, durch Regulari sierung werden die Grenzen gegebenenfalls geglättet, und kleine Regionen werden, falls sie durch Fehldetektionen zustandegekommen sind, entfernt.

Change detection and background extraction by linear algebra

“…Examples of such region based methods are the Shading Model, the Statistical Change Detection method, and the Derivative Model proposed by Skifstad and Jain [8], Aach et al [10], and Hsu et al [9], respectively. These three state-of the art methods will be tested and will serve as a benchmark for the proposed change detection algorithms.…”

“…a shading coefficient . Thus, the intensity is calculated for each point (10) This model was first introduced by Oppenheim et al [26] and is commonly used in computer graphics.…”

“…Given the hypothesis , the threshold can be determined from the following significance test The significance value is equivalent to the probability that exceeds the threshold , given the hypothesis that is due to noise only and not a significant change. For a fixed false alarm rate , the threshold can be obtained from tables of the distribution [10], [27]. Typical values for the significance level are given in Table 11, image noise varies around ten.…”

“…In Section IV, state-of-the art models from Skifstad [8], Hsu [9], and Aach [10] are presented. These models can be categorized as follows.…”

Specialized Video and Physiological Data Coding System for Remote Monitoring

“…A typical example is shown in Fig. 2, where the results of our method and a classic method (by Aach [3]) are compared.…”