<p>Although Transformer-based image super-resolution (SR) methods have achieved remarkable reconstruction performance, their high computational cost-mainly arising from the resource-intensive self-attention (SA) mechanism-remains a major obstacle to deployment on resource-constrained devices. To address this issue, we propose an Adaptive Resolution Block (ARB) and construct an efficient image super-resolution network termed Adaptive Resolution Enhancement Network (AREN). Specifically, the ARB explicitly integrates multi-scale gating with global statistical priors to model scale-dependent response patterns, thereby enabling spatial-channel adaptive feature modulation. Since the ARB primarily leverages non-local feature dependencies, we further introduce a Local Enhancement Feed-Forward Network (LEFN) to refine ARB outputs and encode local contextual information. Extensive experiments demonstrate that AREN achieves superior single image super-resolution (SISR) performance compared to other state-of-the-art models with similar parameter scales, effectively balancing reconstruction quality and computational efficiency.</p>