<p><?tk 2?>Image reconstruction-based anomaly detection techniques identify anomalies by measuring the differences between the reconstructed image and the anomalous image. Emerging diffusion models exhibit powerful generative capabilities, but their reconstruction process, which requires hundreds of denoising steps, imposes a substantial computational burden on anomaly detection tasks. To address this challenge, this paper proposes an anomaly image detection method based on a one-step denoising diffusion model with high- and low-frequency information enhanced. The proposed method consists of three main components. Firstly, a pseudo-anomaly generation module is employed to generate pseudo-anomalous samples, providing supervised training for the model. Secondly, a one-step denoising diffusion model is utilized for rapid reconstruction of anomalous images. To mitigate the issues of image quality degradation and potential reconstruction failure due to the one-step denoising process, we propose a high- and low-frequency information enhanced block that improves the noise prediction capability of the U-Net model. Thirdly, a novel segmentation module is proposed to segment and localize the anomalous regions in the image by comparing the differences between the reconstructed and anomalous images. Experimental results demonstrate that the proposed method achieves image-level and pixel-level AUROC scores of 99.6% and 98.9%, respectively, on the MVTec dataset with only one-step denoising. This method outperforms existing diffusion-based approaches and comes close to the performance of state-of-the-art techniques, demonstrating its superiority and potential for application in industrial anomaly detection.</p>