Viirya

"Image Retrieval: Ideas, Influences, and Trends of the New Age"

By RITENDRA DATTA, DHIRAJ JOSHI, JIA LI, and JAMES Z. WANG, 2008

--

This is a comprehensive paper in which the authors review the theoretical and empirical contributions related to image retrieval.

The article consists of several sections. First it discusses critical aspects of the design of real-world image retrieval systems. For these aspects, some key approaches and techniques of the current decade are presented. As the research of CBIR evolves in the decade, new problems and applications related to image retrieval are brought out.

Two intrinsic problems are derived from the nature of CBIR task:

1. How to represent an image mathematically
2. How to calculate the similarity between images

This is two basic problems related to CBIR and we have learnt in the MMAI course.

The first problem introduces feature extraction techniques. Some features we knew are color, texture, concepts.

To enable accurate image retrieval and other applications, different image similarity measures are proposed in recent years. We are introduced L1, L2, K-L divergence, etc.

The authors then describes two important enable techniques for various image retrieval related tasks and applications: clustering and classification.

Although clustering and classification receive less attention during the early years in this field, the need of practical systems and various applications of CBIR motivate researchers to resort the two techniques.