(A) combined edge-based and region growing method for region extraction

Abstract

In recent years, content-based image retrieval has provided tools and methods for effective searching and browsing of large digital picture libraries. A typical content-based image retrieval system allows a user to provide a query image, from which low-level features are extracted and used to find “similar” images in the database. Shape features along with texture and color are known as primitive features in content-based image retrieval systems. However, these features are not generic for all applications. They depend largely on the types of images. For the images, such as buildings, topographical images, and tree leaves, the texture and color features do not help distinguishing them. The shape feature plays an important role in these applications, and image retrieval becomes mainly shape retrieval in these cases. There are two classes of shape features: contour-based and region-based. Contour-based shape features are simple, but they are not as efficient as region-based shape features. Most systems using the region-based shape feature have to extract the regions first. The prior works on region-based systems still have shortcomings. They are complex to implement, particularly with respect to the region extraction, and do not sufficiently use the spatial relationship between regions in the distance model. In this thesis, a region extraction method that is the combination of an edge-based method and a region growing method is proposed to extract regions inside an object. Shape features of these regions are extracted and stored as the low level description of the object. From these features, the modified IRM (Integrated Region Matching) which includes the adjacency relationship of regions is used to compute the distance between images for similarity search. The experimental results show the effectiveness of our region extraction method as well as the modified IRM. We show that the new method outperforms others.