Comparative study of quantum and classical conditional image generators

Quantum machine learning (QML) has garnered significant attention for its potential to outperform classical machine learning. A notable area within QML is quantum generative adversarial networks (QGANs), which serve as the quantum equivalent to classical GANs, commonly used in image processing and generation. However, the content of images generated by existing QGANs is still uncontrollable. In this study, we propose a conditional encoder for quantum generators, enabling controllable generation of two types of images. Experimental results show that the proposed quantum–classical hybrid GAN is capable of generating images with reasonable quality. We further conduct a systematic comparison between quantum and classical generators under comparable parameter budgets. Rather than claiming superiority, our analysis focuses on identifying the parameter scale at which classical models achieve similar image quality. The results indicate that the proposed quantum generator can reach a comparable level of performance with fewer trainable parameters, highlighting its potential advantage in terms of parameter efficiency. In addition, we investigate the impact of quantum noise by introducing controlled perturbations during the generation process. The results suggest that the proposed model maintains stable performance under moderate noise levels, indicating a certain degree of robustness in noisy quantum settings.