Unsupervised Extraction of Salient Regions for Content Based Image Retrieval

Several content based image retrieval systems rely on unsupervised image segmentation. We argue that in this application context global segmentation methods are not generally applicable. We propose an algorithm which combines local and area based information of multidimensional features, such as luminance, color and texture. Importantly, the selection of the regions with representative attributes is parameter free which guarantees its general applicability. The algorithm is easy to extend to other feature types and was used to extract salient regions from images of a large database consisting of outdoor scenes.