Detecting the moving objects in a video sequence using a stationary camera is an important task for many computer vision applications. This paper proposes a background subtraction approach. As first step, the background is initialized using the block-based analysis before being updated in each incoming frame. Our background frame is generated by collecting the blocks background candidates. The block candidate selection is based on probability density function (pdf) computation. After that, we compute the absolute difference between the background frame and each frame of sequence. A noise filter is applied using the Structure/Texture decomposition in order to minimize the noise caused by background subtraction operation. The binary motion mask is formed using an adaptive threshold that was deduced from the weighted mean and variance calculation. To assure the correspondence between the current frame and the background frame, an adaptation of background model in each incoming frame is realized. After comparing results obtained from the proposed method to other existing ones, it was shown that our approach attains a higher degree of efficacy.Key Words: Motion detection, Background subtraction, Background model, Background update, Video surveillance.
IntroductionMotion detection is a paramount study field of computer vision. Its purpose is to extract the moving objects at time t in a video captured using a stationary camera. The motion detection is used for many applications. Among these applications there are video surveillance, human-machine interaction, the recognition of sign language specific to robotics applications and many others.The approaches mainly used for motion detection can be classified into three categories: the time difference methods, the analysis of optical flow, and the background subtraction methods. For the first one, the time difference being the calculation of the difference between two or more consecutive images in order to extract the moving area [1], [2], but the problem in this approach is that the detected objects are incomplete and poorly presented. The second approach is the calculation of the optical flow [3], [4] which provides all information about the movement, but the real-time implementation is difficult and calculation of