Manuscript received June 12, 2025; revised July 24, 2025; accepted October 9, 2025; published 16, 2026.

Abstract—A key tool for organizing and extracting visual information from large picture databases is Content-Based Image Retrieval (CBIR). To make CBIR systems much more accurate and efficient, this research proposes a new combined approach that uses deep feature extraction along with Quantum Grey Wolf Optimization (QGWO). The suggested system captures complex visual patterns across a variety of images categories by utilizing pre-trained Convolutional Neural Networks (CNNs) for reliable and advanced feature extraction. This paper presents the retrieval of single objects and multi-objects. The deep learning techniques utilized in this work are Inception V1, Inception V2, and ResNet50, chosen as they represent progressively advanced CNN architectures with increasing depth and feature extraction capability. To make the process faster and more effective, these features are improved using QGWO, a method that combines ideas from quantum computing with the social behaviour and hunting strategies of grey wolves. Our combined method performs better than existing CBIR algorithms in precision, recall, accuracy, and F1-Scores, based on thorough testing with the Corel-1K, Corel-5K, Corel-10K image datasets. The fusion of deep learning with a quantum-inspired optimization approach resulted in a retrieval accuracy of 99.20%, and percentage of wrongly retrieved images obtained is 1.08% using Corel-1k. The accuracy achieved using Corel-5k is 99.12% and Corel- 10K is 99.04%. The results prove the computational efficiency of contemporary CBIR system.

Keywords—Content-Based Image Retrieval (CBIR), Convolutional Neural Networks (CNNs), Inception V1, Inception V2, ResNet50, Quantum Grey Wolf Optimization (QGWO)

Cite: Sarva N. Kumar and Ch S. Kumar, "A Hybrid Deep Feature Extraction Framework with Quantum Grey Wolf Optimization for Enhanced Content-Based Image Retrieval," Journal of Image and Graphics, Vol. 14, No. 1, pp. 38-48, 2026.

Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC-BY-4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.