Logo image
Back
Quantum Convolutional Neural Network for Image Retrieval
Conference proceeding

Quantum Convolutional Neural Network for Image Retrieval

Fahimeh Alaei and Alireza Alaei
Pattern Recognition and Computer Vision, Vol.II, pp.203-214
Lecture Notes in Computer Science
8th Asian Conference on Pattern Recognition, ACPR 2025, 8th (Gold Coast, Australia, 10/11/2025–13/11/2025)
09/11/2025
DOI: https://doi.org/10.1007/978-981-95-4398-4_15
Appears in  Recent Faculty of Science and Engineering Publications

Metrics

4 Record Views

Abstract

Feature Extraction Image Retrieval Quantum Computing Quantum Convolutional Neural Network
In the era of information overload, retrieving relevant images efficiently from large-scale databases is critical in different fields, such as document analysis, medical imaging, remote sensing, and surveillance. This paper introduces a novel image retrieval model called the Hybrid Quantum Neural Networks (HQNN). The proposed model embeds parameterised quantum circuits within a classical CNN architecture to enhance feature representation and reduce parameter overhead. Compact features extracted from convolutional layers are processed through parallel quantum layers to capture high-level feature abstractions and obtain more accurate results. Various similarity measures, including Quantum, Euclidean, and Manhattan, are employed to achieve final retrieval results. We evaluated the proposed model on the Corel-5K dataset, and the results revealed that HQNN outperformed its classical CNN counterpart. Notably, HQNN achieved a Top-20 precision of 85.57% with quantum similarity, surpassing the classical CNN’s performance. These findings validate the potential and suitability of quantum computing in content-based image retrieval (CBIR). This work contributes to the growing field of Quantum Machine Learning (QML), offering a promising direction for next-generation intelligent image retrieval systems.

Details

Logo image