Dynamic Background Subtraction Using Texton Co-occurrence Matrix

Abstract

Moving object detection in the presence of changing illumination and non-stationary background such as swaying of trees, fountains, ripples in water, flag fluttering in the wind, camera jitters, noise, etc., is known to be very difficult and challenging task. Background subtraction (BS) is the most sought after technique for moving object detection. Still, most of the BS techniques do not take into account the spatial relationship between the pixels. In this paper, we have presented a novel BS algorithm using the properties of texton co-occurrence matrix (TCM) for accurately detecting the moving objects. TCM is a popular technique in the field of image retrieval. However, its adoption in BS is not reported in the literature. TCM integrates the colour, texture, and shape features in background modelling. It is computed in a neighbourhood region of the pixel. This implicitly utilizes image features and the spatial relationship between the pixels in the BS. Quantitative and qualitative results of the proposed algorithm is shown with state-of-the-art techniques, to prove its efficacy for moving object detection in presence of dynamic backgrounds.