Mathphys-Guided Coarse-to-Fine Anomaly Synthesis with SQE-Driven Bi-Level Optimization for Anomaly Detection

Currently, industrial anomaly detection suffers from two bottlenecks: (i) the rarity of real-world defect images and (ii) the opacity of sample quality when synthetic data are used. Existing synthetic strategies overlook the underlying physical causes of defects, leading to inconsistent, low-fidelity anomalies that hamper model generalization to real-world complexities. In this paper, we introduce a novel and lightweight pipeline that generates synthetic anomalies through MathPhys model guidance of the typical defect mechanisms to produce realistic anomaly masks, refines them via a Coarse-to-Fine approach to enforce global PDE consistency (npcF) and then restores high-frequency details and improves local fidelity (npcF++) and employs a bi-level optimization strategy with a Synthesis Quality Estimator (SQE) ensuring that high-quality synthetic samples receive greater emphasis during training. Experiments demonstrate that our method achieves state-of-the-art results in both image-/pixel-AUROC.