Expert image retrieval system using directional local motif XoR patterns

• New methodology of motif representation is presented using 1 × 3 grids.
• New motif (1 × 3 grids) is able to collect four directional information.
• XoR operation is applied on both proposed and existing motif representation.
• Three experiments conducted on Corel-5000, Corel-10000 and MIT-VisTex databases.
• Proposed method shows a significant improvement in terms of ARP and ARR.

This paper presents a new image feature descriptor, namely directional local motif XoR patterns (DLMXoRPs) for image retrieval application. The proposed motif representation is entirely different from existing motif. The DLMXoRP presents a novel technique for the calculation of motif using 1 × 3 grids. The proposed motif (1 × 3) representation is having a flexible structure; hence it can able to extract all directional information. This flexibility is not present in the existing (2 × 2) motif. Further, the XoR operation is performed on the transformed new motif images which are not present in the literature (local binary patterns (LBP) and motif co-occurrence matrix (MCM)). To elevate the benefits of DLMXoRP, we compare it with the Motif XoR pattern (MXoR) which is calculated by applying the XoR operation on existing transformed motif image. The performance of the proposed method is tested by conducting three experiments on Corel-5000, Corel-10000 and MIT-VisTex databases. The results after investigation show a significant improvement in terms of average retrieval precision (ARP) and average retrieval rate (ARR) as compared to the state-of-the-art techniques for image retrieval.