Improving the Graph-Based Image Segmentation Method

Sensor devices are widely used for monitoring purposes. Image mining techniques are commonly employed to extract useful knowledge from the image sequences taken by sensor devices. Image segmentation is the first step of image mining. Due to the limited resources of the sensor devices, we need time and space efficient methods of image segmentation. In this paper, we propose an improvement to the graph-based image segmentation method already described in the literature and considered as the most effective method with satisfactory segmentation results. This is the preprocessing step of our online image mining approach. We contribute to the method by re-defining the internal difference used to define the property of the components and the threshold function, which is the key element to determine the size of the components. The conducted experiments demonstrate the efficiency and effectiveness of the adjusted method