Reduced-reference metric design for objective perceptual quality assessment in wireless imaging

摘要: The rapid growth of third and development future generation mobile systems has led to an increase in the demand for image video services. However, hostile nature wireless channel makes deployment such services much more challenging, as case a wireline system. In this context, importance taking care user satisfaction with service provisioning whole been recognized. related user-oriented quality concepts cover end-to-end subjective factors experiences service. To monitor adapt system resources, performance indicators that represent integrity have be selected objective measures correlate well perceived by humans. Such perceptual metrics can then utilized optimize perception associated applications technical systems. paper, we focus on design reduced-reference use imaging. Specifically, normalized hybrid metric (NHIQM) relevance weighted L"p-norm are designed. main idea behind both feature-based relates fact human visual (HVS) is trained extract structural information from viewing area. Accordingly, NHIQM designed account different artifacts observed our distortion model link. extent which individual present given obtained measuring features. overall measure computed weighting sum features respective weight experiments. proposed differ mainly pooling amount produced. While performs at transmitter produce single value reduced-reference, requires all involved feature values transmitted received perform differences receiver. addition, non-linear mapping functions developed relate predicted mean opinion scores (MOS) saturations HVS. evaluation prediction reveals their excellent correlation terms accuracy, monotonicity, consistency. This holds not only images taken training but also generalization unknown images. it shown approach outperform other prominent performance.