Article: Lattice vector quantisation for indexing and retrieval of medical images using texture features based on 2-D Wold decomposition Journal: International Journal of Image Mining (IJIM) 2015 Vol.1 No.1 pp.23 - 44 Abstract: Medical image retrieval to search for clinically relevant and visually similar images to support radiologists' examination has been attracting lot of research interest. It provides a baseline essential for confirmation, comparison and evaluation of suspicious radiographic signs detected on current examination images. Content-based image retrieval (CBIR) is an important alternate and complement to traditional text-based image retrieval (TBIR) using keywords. The proposed CBIR system is based on the effective use of texture information within images using 2-D Wold decomposition of homogeneous random fields. To speed up the search process, selected features are extracted and indexed using the proposed indexing structure. This paper describes a new efficient indexing structure for elongated vectors of feature space using lattice vector quantisation (LVQ). Selected features are indexed based on norm, leader and permutation indices. Simple Euclidean distance measure is used to display the retrieved images in the decreasing order of similarity. Features for harmonic structures and evanescent components are compared based on precision and recall. Experiments were carried out on MR-T2 axial brain slices. Features of harmonic structure gives better results when compared to evanescent components. Inderscience Publishers - linking academia, business and industry through research

Title: Lattice vector quantisation for indexing and retrieval of medical images using texture features based on 2-D Wold decomposition

Authors: A.N. Krishna; B.G. Prasad

Addresses: Department of Computer Science and Engineering, S.J.B. Institute of Technology, Bangalore-560060, India ' Department of Computer Science and Engineering, B.N.M. Institute of Technology, Bangalore-560070, India

Abstract: Medical image retrieval to search for clinically relevant and visually similar images to support radiologists' examination has been attracting lot of research interest. It provides a baseline essential for confirmation, comparison and evaluation of suspicious radiographic signs detected on current examination images. Content-based image retrieval (CBIR) is an important alternate and complement to traditional text-based image retrieval (TBIR) using keywords. The proposed CBIR system is based on the effective use of texture information within images using 2-D Wold decomposition of homogeneous random fields. To speed up the search process, selected features are extracted and indexed using the proposed indexing structure. This paper describes a new efficient indexing structure for elongated vectors of feature space using lattice vector quantisation (LVQ). Selected features are indexed based on norm, leader and permutation indices. Simple Euclidean distance measure is used to display the retrieved images in the decreasing order of similarity. Features for harmonic structures and evanescent components are compared based on precision and recall. Experiments were carried out on MR-T2 axial brain slices. Features of harmonic structure gives better results when compared to evanescent components.

Keywords: content-based image retrieval; CBIR; Wold features; lattice vector quantisation; LVQ; indexing structure; image indexing; medical images; texture features; Wold decomposition; brain imaging.

DOI: 10.1504/IJIM.2015.070023

International Journal of Image Mining, 2015 Vol.1 No.1, pp.23 - 44

Received: 08 Aug 2014
Accepted: 13 Feb 2015
Published online: 24 Jun 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article

Title: Lattice vector quantisation for indexing and retrieval of medical images using texture features based on 2-D Wold decomposition

Keep up-to-date