Principal Objects Detection Using Graph-Based Segmentation and Normalized Histogram

In this paper, we introduce a new method to distinguish the principal objects in image datasets using graph-based segmentation and normalized histogram (PODSH). Unlike the usual object detection systems which require the input objects, we propose a new approach to recognize objects one might focus on when taking images. Motivated by the habit of taking picture, we suppose that the position of a main object is located near the image centre and this object always holds a large area. The normalized histogram is added to increase the effect of our system. In the experiment, we used images which consist of objects to test the precision of PODSH. Our system is implemented by Matlab.