Correcting Deviations from Normality: A Reformulated Diffusion Model for Multi-Class Unsupervised Anomaly Detection

The advanced image generation capabilities of recent diffusion models have spurred research into their application for Reconstruction-based unsupervised anomaly detection. However, these methods may struggle with maintaining pixel-level structural integrity and recovering the anomaly-free content of abnormal regions, especially in multi-class scenarios. Furthermore, diffusion models are inherently designed to generate images from pure noise and struggle to selectively alter anomalous regions in an image while preserving normal ones. This leads to potential degradation of normal regions during the reconstruction process, hampering the effectiveness of anomaly detection. This paper introduces a reformulation of the standard diffusion model geared toward selective region alteration, allowing the accurate identification of anomalies.Our proposed Deviation correction diffusion (DeCo-Diff) model preserves the normal regions and encourages transformations exclusively on anomalous areas. By modeling anomalies as noise in the latent space, our method leverages the learned distribution of normal images to accurately reconstruct normal regions while altering only the anomalous areas. This selective approach enhances the reconstruction quality, facilitating effective unsupervised detection and localization of anomaly regions. Comprehensive evaluations demonstrate the superiority of our method in accurately identifying and localizing anomalies in complex images, with pixel-level AUPRC improvements of 11-14% over state-of-the-art models on well-known anomaly detection datasets. The code is available at https://anonymous.4open.science/r/DeCo-Diff