Local peak valley co-occurrence patterns: a new feature descriptor for image retrieval

Abstract

In this paper, we propose a novel feature extraction and retrieval technique for medical images. Our proposed technique (LPVCoP) extracts similarity between grayscale images by using the relationship between reference pixel and its surrounding neighbor pixels through peak/valley edges which are obtained by taking directional derivatives. LPVCoP uses the co-occurrence among the first-order directional derivatives of reference pixel and second neighborhood adjoining pixels. The working of the proposed technique has been checked by implementing it in MATLAB and verifying the result on Computer Tomography (CT) databases. The retrieval results of the proposed technique are compared with the existing technique of image retrieval in terms of average retrieval rate (ARR) and average retrieval precision (ARP). The retrieval results of the LPVCoP are better as compared to the existing techniquesfor image retrieval.