SR-Diff: semantic-guided diffusion for high-fidelity image restoration | Connection Science | DeepDyve