<p>Facial Expression Analysis is essential for emotion-aware technologies such as assistive interfaces, behavioral monitoring, and intelligent surveillance, where low-latency processing is required. This work introduces a hybrid quantum-classical framework addressing key challenges in landmark precision, class imbalance, and hardware feasibility. First, a novel method is proposed for correcting facial landmark positions using edge-aware scanning guided by quantum-derived probabilities, improving spatial alignment and clarity of expressive regions. Second, quantum distance measurement uses confidence-interval–guided adaptive shot selection to stabilize estimates under noise while preserving low circuit depth. Third, quantum-computed distances are transformed into enriched representations as classifier training data, capturing relative similarity, confidence, and distributional structure to help mitigate class imbalance and improve emotion category separability. The enriched features are processed by a neural network for emotion recognition. Evaluations on diverse datasets show improvements in accuracy, generalization, robustness, and real-time feasibility while maintaining compatibility with noisy intermediate-scale quantum hardware.</p>