A high-performance content-based image retrieval system on PC cluster

Image Retrieval problem is concerned with retrieving images that are relevant to user's requests from an image database. Traditional information retrieval techniques, such as text-based information retrieval, have - ignored the most remarkable characteristics of real-world objects. To alleviate the arduous job in getting the important image contents, Charlet E. Jacobs [1] has facilitated wavelet transform. [2] and some other image processing techniques to design a CBIR (content-based image retrieval) system in 1995. Various methods of CBIR systems suffer two main drawbacks, retrieval inefficiency and use of single similarity metric. In this paper, we propose a novel idea in designing a wavelet-based CBIR system which makes use of mesh-of-trees architecture to overcome the above problems. The design is sketched and implemented on a PC cluster environment for its superior performance and scalability.