Radial Moments for Image Retrieval

Introduction

Images have always been considered an effective medium for presenting visual data in numerous applications ranging from industry to academia. With the development in technology, a large amount of images are being generated everyday. Therefore, managing and indexing of images become essential in order to retrieve relevant images effectively and efficiently. Early work on image retrieval can be traced back to 1970s. In 1979, a conference was held regarding database techniques for pictorial applications (Blaser, 1979). Since then, the research pertaining to image database management has influenced several researchers (Chang and Fu, 1979; Chang and Fu, 1980; Chang and Kunii, 1981; Chang et al., 1988). In traditional systems, textual annotations of images were used to describe images. Afterwards, the images were searched using text based approach from traditional database management system such as SQL query. A comprehensive survey of text based retrieval can be found in (Chang and Hsu, 1988; Chang et al. 1988). In text based image retrieval, images are organized by semantic hierarchies to facilitate navigation and browsing using standard keyword queries. However, the textual annotation of images is a cumbersome task, which requires intensive manual labor for large image databases. Apart from that, the use of keywords become unwieldy, which do not meet human visual perception. Consequently, it is intricate for the traditional text based methods to support wide range of image databases and task dependent queries. Humans are capable of identifying an extensive range of objects irrespective of their size and orientation. A widespread research has been pursuing to develop an automated system to imitate this basic capability of human visual perception. In 1992, the National Science Foundation of United States organized a workshop on visual information management system (Jain, 1991) to discover new trends in image database management systems. It was discussed that a more efficient and spontaneous way to represent and index visual information would be based on attributes that are inherent in images themselves. In other words, the representation and indexing of images should be based on their actual visual content rather than textual annotations. Since then, the research on content based image retrieval (CBIR) has been developing rapidly (Cawkill, 1993; Dowe, 1993; Jain et al., 1995; Mejdoub et al., 2009, Arnold et al., 2000; Lew, 2006). CBIR has several applications such as medical imaging, trademark matching, digital library, computer aided design, military services, crime prevention, target detection, architectural and engineering design, geographic information, chromosome identification, surveillance tasks, etc.