Title: An automatic image retrieval system using multi-scale local ternary pattern

Authors: Megha Agarwal

Addresses: Department of Electronics and Communication Engineering, Jaypee Institute of Information Technology, Noida, Uttar Pradesh, India

Abstract: Content-Based Image Retrieval (CBIR) addresses the issue of finding out the relevant images automatically from the vast image repositories. Images are compared based on the extracted features and, hence, the feature extraction is highly responsible in CBIR system performance. In this paper, a unique feature, Multi-Scale Local Ternary Pattern (MSLTP) is designed. Mostly, local patterns consider original images for feature extraction but in MSLTP, the images are analysed in five different scales and, hence, all the image information, whether fine or coarse is captured in feature extraction. Along with this, texture information is computed by taking small neighbourhoods and comparing surrounding pixels. The pattern of intensity variation in local neighbourhoods is captured. Performance of the system is validated on very distinct benchmark data sets Corel 1K and MIT VisTex. Significant improvement is observed in terms of retrieval precision and recall, as compared to the other handcrafted features available in the literature.

Keywords: image retrieval; local pattern; Corel 1K; MIT texture.

DOI: 10.1504/IJCAT.2023.133296

International Journal of Computer Applications in Technology, 2023 Vol.72 No.3, pp.181 - 188

Received: 08 Sep 2022
Received in revised form: 05 Nov 2022
Accepted: 24 Dec 2022
Published online: 11 Sep 2023 *

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