Quantum image processing relies on the efficient encoding of classical image data into quantum states, yet the practical performance of existing encoding schemes on noisy intermediate-scale quantum (NISQ) hardware remains insufficiently understood. Most prior studies have emphasized asymptotic resource scaling or idealized circuit constructions, leaving a gap between theoretical proposals and realistic device performance. In this work, we present a systematic, hardware-realistic benchmark of six representative quantum image encoding schemes—Basis, Angle, Amplitude, Flexible Representation of Quantum Images (FRQI), Novel Enhanced Quantum Representation (NEQR), and Order-Encoded Quantum Image Model (OQIM)—using image retrieval as a test task. Our analysis explicitly incorporates transpilation overhead, device connectivity constraints, and calibrated noise models derived from multiple IBM quantum processors. Reconstruction fidelity is evaluated across different image sizes, circuit depths, and measurement budgets. We find that qubit efficiency alone is a poor predictor of practical performance on NISQ devices. Instead, post-transpilation circuit depth emerges as the dominant factor governing noise resilience and information preservation. Encoding schemes with favorable asymptotic qubit scaling but deep circuit implementations suffer severe degradation under realistic noise, while shallower constructions retain meaningful image structure. Among the considered schemes, amplitude encoding consistently achieves the highest reconstruction fidelity under hardware-calibrated simulation; angle encoding, validated on physical hardware, demonstrates high fidelity. These results clarify the practical performance hierarchy of quantum image encodings under NISQ conditions and provide experimentally grounded guidance for encoding selection and circuit design on near-term quantum platforms.

Close

PrevNext

View figure in article

• Quantum information processing