Hierarchical graph-based segmentation in detection of object-related regions

Abstract

Object detection is an important task in computer vision. Recently, several unsupervised approaches have been proposed to cope with this problem in a category-independent manner. This work evaluates the adoption of a hierarchical graph-based segmentation along with an state-of-the-art method to detect object-related regions. A hierarchical segmentation approach produces a set of partitions at different detail levels, in a way that a coarser level can be obtained by a simple merge of finer ones. Experimental results show that our proposal obtains an increase of 11% in object detection rate.