Complementary semantic model for content-based image retrieval

Abstract

Multimedia technology is known as the single repository source for information retrieval that includes images, audio and video. Nowadays, it is growing with tremendous amount in the field of communication technology. Traditional text based retrieval techniques of information retrieval are inefficient at searching user's intended information from the large multimedia database. To address this problem, content based image retrieval techniques are applied for information retrieval. Content-based image retrieval has been hot topic from past two decades by many researchers' community. This topic covers many fields such as image processing, artificial intelligence, pattern recognition, and data mining and beside other related area. In this paper image retrieval techniques on the basis of color, texture and shape are explained, studies and compared. While relevance feedback technique, system accuracy, efficiency have also been discussed. Emphasize on the concept of 'semantic-based image retrieval techniques' have been proposed in this paper, which is better technique to overcome the problem of 'semantic gap' in the context of image retrieval.

A concept-based model for image retrieval systems

“…Although low-level features can describe the content of simple images, they cannot properly describe complex images, containing high-level concepts. This is the main challenge of CBIR systems and is called ''semantic gap'' in the literatures [3][4][5][6][7]. In addition to this challenge, CBIR systems usually ignore the relations and context among the image objects in the retrieval process.…”

“…Then, the query image is described by set of low-level features such as color, texture and shape. Finally, the most similar images according to the extracted features are retrieved [1][2][3]. Although low-level features can describe the content of simple images, they cannot properly describe complex images, containing high-level concepts.…”