A Robust Texture-Based Background Subtraction Algorithm for Moving Object Detection in Video Sequences

In this paper, we propose a robust texture-based background subtraction algorithm for moving object detection in video sequences. A robust texture descriptor so-called local fuzzy pattern (LFP) histogram is proposed and employed to initialize and maintain the background models of each pixel in a video frame. The calculated LFP histogram of each pixel in the new frame is compared to its corresponding background models for classifying the pixel into the class of foreground or moving objects. Compared with the other approach, experimental results show that our approach possesses better adaption and tolerance abilities for the conditions of shadows and illumination variation.