Peak Valley Edge Patterns: A New Descriptor for Biomedical Image Indexing and Retrieval

Subrahmanyam Murala, Jonathan Q.M. Wu; Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2013, pp. 444-449

Abstract

A new algorithm meant for biomedical image retrieval application is presented in this paper. The local region of image is represented by peak valley edge patterns (PVEP), which are calculated by the first-order derivatives in 0?, 45?, 90? and 135? directions. The PVEP differs from the existing local binary pattern (LBP) in a manner that it extracts the directional edge information based on firstorder derivative in an image. Further, the effectiveness of our algorithm is confirmed by combining it with Gabor transform. The performance of the proposed method is tested on VIA/I-ELCAP database which includes region of interest computer tomography (ROI-CT) images. Performance analysis shows that the proposed method improves retrieval results from 79.21% to 86.13% and 51.91% to 55.06% as compared to LBP in terms of average precision when number of top matches considered is 10 and 100 respectively.

Related Material

[pdf]
[bibtex]
@InProceedings{Murala_2013_CVPR_Workshops,
author = {Murala, Subrahmanyam and Wu, Jonathan Q.M.},
title = {Peak Valley Edge Patterns: A New Descriptor for Biomedical Image Indexing and Retrieval},
booktitle = {Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
month = {June},
year = {2013}
}