IJRECE VOL. 6 ISSUE 3 ( JULY - SEPTEMBER 2018) ISSN: 2393-9028 (PRINT) | ISSN: 2348-2281 (ONLINE) 2131 | Page INTERNATIONAL JOURNAL OF RESEARCH IN ELECTRONICS AND COMPUTER ENGINEERING A UNIT OF I2OR A Proposed Method for Background Subtraction and adding Scribbles in Video Matting Mervat El-Seddek 1 , Hossam El-Din Moustafa 2 1 Electronics and Communications Department - Misr Engineering and Technology Institute 2 Electronics and Communications Department – Faculty of Engineering – Mansoura University (E-mail: Hossam_moustafa@hotmail.com) Abstract: Identifying moving objects from a video sequence is a fundamental task in most computer-vision applications. It is essential for visual tracking systems. In this paper, a new algorithm for background subtraction has been proposed and evaluated. Decision-level fusion of three background subtraction algorithms; frame difference, approximated median, and Mixture of Gaussians; has been utilized. It has been found that the percentage error for each frame has reduced remarkably when applying the proposed technique.Moreover, an entirely-automatic algorithm for adding scribbles in video matting is proposed to improve video matting results depending on the suggested background subtraction algorithm. The overall performance is evaluated. The average error has been reduced effectively and the overall performance has been enhanced. Keywords-Background subtraction; frame difference; mixture of Gaussians; approximated median scribbles; video matting I. INTRODUCTION Foreground detection in a video sequence is one of the most important tasks in video surveillanceand tracking systems [1]. Background subtraction is utilized for discriminating moving objects from the background scene. In background subtraction, the current image is subtracted from a reference one. This leaves new objects in the scene [2]. Fig. (1) shows the four maincomponents of a background subtraction system. These are preprocessing, background modeling, foreground detection, and data validation [3]. The simplest background modeling algorithm is frame differencing [4]. It uses the video frame at time (t –1) as the background model for the frame at time t. Frame differencing performs poorly if the background is not truly static (e.g., fluttering leaves, water, waves). The generalized Mixture of Gaussians (MoG) has been utilized to model complex, non-static backgrounds [2]. However, backgrounds with fast variations cannot be accurately modeled with just a few Gaussians, causing problems for sensitive detection. Fig. 1. Basic Steps for Background Subtraction Algorithms Median filtering is one of the most commonly-used backgrounds modeling techniques [5], [6], [7]. The background estimate is defined to be the median at each pixel location of all frames in the buffer.The assumption is that the pixel stays in the background for more than half of the frames in the buffer. The approximated median filter was proposed such that the running estimate of the median is incremented by one if the input pixel is larger than the estimate, and decreased by one if small [8]. The present paper proposes a new combinational method for background modeling. It relies on introducing some modification in each of the three previously explained techniques and then merging the outputs of the three techniques to obtain the output of the proposed fusion technique. The organization of the paper is as follows: Section 2 presents the proposed algorithm and a comparative study of its performance with the three traditional methods. Section 3 introduces the matting problem. In section 4, the proposed method for adding scribbles is presented. Section 5 is the conclusion. II. THE PROPOSED ALGORITHM The proposedtechnique relies on two consecutive steps. The former utilizes image processing techniques to enhance results of each single background subtraction method.The most important step is the second one where data fusion techniques