I.J. Image, Graphics and Signal Processing, 2016, 7, 41-48 Published Online July 2016 in MECS (http://www.mecs-press.org/) DOI: 10.5815/ijigsp.2016.07.05 Copyright © 2016 MECS I.J. Image, Graphics and Signal Processing, 2016, 7, 41-48 Motion Segmentation from Surveillance Video using modified Hotelling's T-Square Statistics Chandrajit M 1 Maharaja Research Foundation, MIT, Mysore, India E-mail: chandrajith.m@gmail.com Girisha R 2 , Vasudev T 1 PET Research Foundation, PESCE, Mandya, India E-mail: write2girisha@gmail.com, vasu@mitmysore.in Abstract—Motion segmentation is an important task in video surveillance and in many high-level vision applications. This paper proposes two generic methods for motion segmentation from surveillance video sequences captured from different kinds of sensors like aerial, Pan Tilt and Zoom (PTZ), thermal and night vision. Motion segmentation is achieved by employing Hotelling's T-Square test on the spatial neighborhood RGB color intensity values of each pixel in two successive temporal frames. Further, a modified version of Hotelling's T-Square test is also proposed to achieve motion segmentation. On comparison with Hotelling's T- Square test, the result obtained by the modified formula is better with respect to computational time and quality of the output. Experiments along with the qualitative and quantitative comparison with existing method have been carried out on the standard IEEE PETS (2006, 2009 and 2013) and IEEE Change Detection (2014) dataset to demonstrate the efficacy of the proposed method in the dynamic environment and the results obtained are encouraging. Index Terms—Motion segmentation,Video surveillance, Spatio-temporal, Hotelling's T-Square test. I. INTRODUCTION Video surveillance has become one of the most active areas of research in computer vision. Generally, video surveillance system involves activities like motion segmentation, object classification, object recognition, object tracking and motion analysis. Moving object segmentation is extracting the regions of the video frame which are non-stationary. Object classification is classifying the objects such as a person, vehicle or animal. Identifying the object of interest is object recognition. Motion tracking is establishing frame by frame correspondence of the moving object in the video sequence. Finally, analyzing the object motion and interpretation leads to motion analysis. Motion segmentation is a vital task in video surveillance as the subsequent tasks of the video surveillance system are dependent on the accurate output of motion segmentation. The surveillance video sequences are generally captured through different sensors like aerial, PTZ, thermal and night vision. The captured sequence consists of noise and illumination variations, which makes the motion segmentation from surveillance videos a challenging task [35, 36, 38]. Therefore, the research focuses on developing efficient and reliable motion segmentation algorithm to overcome the mentioned limitations and to extract foreground information from the image data for further analysis. Several techniques are proposed in the literature for motion segmentation and these techniques can be categorized as conventional background subtraction [14], statistical background subtraction [2, 10, 21, 28, 30, 32, 34, 37], temporal differencing [1, 5, 6], optical flow [9] and hybrid approaches [3, 4, 7, 8, 15, 16, 17, 31, 33, 40]. The conventional background subtraction technique initially builds the background model and the new frame is subtracted from the background model for motion segmentation. The statistical background subtraction technique builds the background model by using individual pixel or group of pixels dynamically and then each pixel from the current frame is treated as foreground or background by comparing the statistics of the current background model. In the temporal difference method, absolute difference of successive frames is done to segment motion pixels. Optical flow technique computes the flow vectors of every pixel and then segments the moving object. The hybrid techniques use the combination of above techniques for segmentation of moving objects in video sequences [11]. A brief review of the existing works is discussed in the subsequent section. This paper proposes two generic methods for segmenting moving objects from surveillance video in the dynamic environment by fusing spatial neighborhood information from color video frames in a temporal statistical framework. The article is organized as follows. Section 2 reviews the related works on segmentation methodologies. The overview of the proposed work is described in Section 3. Section 4 elaborates the proposed work. The experimental results and conclusion are reported in Section 5 and 6 respectively.