Fine-grained tampered area detection method in document images based on multi-transformation domain attention fusion

With the continuous evolution of document forgery technology, high-quality tampering poses severe challenges to traditional detection methods based on single-domain statistics and edge structures. Therefore, the study proposes a method based on multi-transformation domain attention fusion and fine-grained artifact enhancement to detect fine-grained tampered areas in document images. Through seven-domain artifact generation, domain-level and inter-domain attention weighting, multi-scale dilated convolution spatial enhancement, and pixel-level, region-level and pixel-level joint loss constraints, the tampered area can be accurately located. Experimental results showed that this method still maintained a precision of 79.3% and a recall of 63.7% in a scene with an artifact ratio of 1%. In a multi-type tampering scene, the lowest balanced error rate was about 10.0%. In the noise interference experiment, the pixel accuracy was above 96%. It obtained a Matthews correlation coefficient of 86.9% and a balanced accuracy of 91.3% in the ablation experiment. The results show that this detection method can significantly improve the detection capability and robustness of weak artifacts in complex environments, and provide theoretical and methodological support for the subsequent development of intelligent document forensic systems.