Accurate and Lightweight Image Super-Resolution With Model-Guided Deep Unfolding Network

Deep neural networks (DNNs) based methods have achieved great success in single image super-resolution (SISR). However, existing state-of-the-art SISR techniques are designed like black boxes lacking transparency and interpretability. Moreover, the improvement visual quality is often at price of increased model complexity due to black-box design. In this paper, we present advocate an explainable approach toward named model-guided deep unfolding network (MoG-DUN). Targeting breaking coherence barrier, opt work with a well-established prior nonlocal auto-regressive use it guide our DNN By integrating denoising regularization as trainable modules within learning framework, can unfold iterative process model-based into multi-stage concatenation building blocks three interconnected (denoising, nonlocal-AR, reconstruction). The design all leverages latest advances including dense/skip connections well fast implementation. addition explainability, MoG-DUN accurate (producing fewer aliasing artifacts), computationally efficient (with reduced parameters), versatile (capable handling multiple degradations). superiority proposed method SR RCAN, SRMDNF, SRFBN substantiated by extensive experiments on several popular datasets various degradation scenarios.