Content-based image retrieval by shape matching

Successful retrieval of images from large-scale image collections is one of the current problems in the field of multimedia digital libraries. In this context, the use of textual descriptions to represent and query images may provide poor results, due to subjectivity of descriptions. Conversely, image retrieval based on visual content of images has revealed more powerful. In this paper, we present a system, which combines different techniques from computer vision to perform content-based image retrieval based on shape matching. The architecture of the system is portrayed and algorithms underpinning the developed Matlab toolbox are briefly described. As an application example, the proposed system is employed to retrieve images in a database of real-world pictorial images

I. Introduction

The growing emergence of image databases employed nowadays in several application areas, such as entertainment, art, medicine and industry, makes image retrieval a challenging problem to be addressed in the development of multimedia digital libraries. In this context, the use of textual descriptions to represent and query visual data may provide results which could not match at all user's expectation. Conversely, image retrieval based on visual content of images has revealed more powerful [1]. Indeed, content-based image retrieval exploits primitive visual features that humans use to analyze and understand the content of an image, thus reducing the cognitive effort of the user in accessing the database and retrieving results that better match user's expectation. Several approaches have appeared in the literature which perform visual querying by example taking into account different facets of pictorial data to express the image contents, such as colour [2], object shape [3], texture [4], or a combination of them [5], [6]. Among these, search by matching shapes of portions of images is a natural way to pose a query in image databases.