This paper addresses the challenge of image restoration using diffusion models by proposing a novel method that decouples the reverse sampling process from data consistency steps. The proposed method alternates between a reconstruction phase to maintain data consistency and a refinement phase that enforces the prior via diffusion purification. This approach improves efficiency and versatility, making it adaptable for various image restoration tasks such as denoising, deblurring, inpainting, and super-resolution. The method is validated through comprehensive experiments, demonstrating superior performance compared to existing techniques in terms of reconstruction quality and computational efficiency. The paper also discusses the adaptability of the method to accelerated samplers and latent diffusion models, further enhancing its practical utility.This paper addresses the challenge of image restoration using diffusion models by proposing a novel method that decouples the reverse sampling process from data consistency steps. The proposed method alternates between a reconstruction phase to maintain data consistency and a refinement phase that enforces the prior via diffusion purification. This approach improves efficiency and versatility, making it adaptable for various image restoration tasks such as denoising, deblurring, inpainting, and super-resolution. The method is validated through comprehensive experiments, demonstrating superior performance compared to existing techniques in terms of reconstruction quality and computational efficiency. The paper also discusses the adaptability of the method to accelerated samplers and latent diffusion models, further enhancing its practical utility.