Motion segmentation from surveillance videos using T-test statistics

Abstract

Motion segmentation is an important task in video surveillance and in many high level vision applications. In this paper, an adaptive method using statistics in temporal framework to segment moving objects from surveillance video sequences captured in dynamic environment is proposed. The proposed method first preprocesses the input frames of video using Gaussian filter for noise reduction. Motion segmentation is done by employing statistical T-test on neighborhood RGB color intensity values of each pixel in two successive temporal frames. Several experiments along with comparison with existing method have been carried out on the IEEE PETS (2009 and 2013) and IEEE Change Detection (2014) datasets which include thermal, normal, PTZ, aerial and night vision sensor videos to demonstrate the efficacy of the proposed methods in dynamic environment and results obtained are encouraging.