基于PVM的Seam Carving篡改检测算法OA

To address the insufficient robustness and generalization of existing Seam Carving tampering detection al-gorithms when dealing with post-processing attacks such as JPEG compression,Gaussian blurring,and Gaussian noise,we propose a Seam Carving tampering detection algorithm based on Pyramid Vision Mamba(PVM).Firstly,we construct a PVM encoder that leverages its multi-scale modeling capability to capture the global contextual infor-mation of tampering traces.Secondly,we design an eight-direction local attention mechanism to detect correlation anomalies caused by seam operations.Finally,a robustness-enhancing regularization term is introduced into the loss function to effectively suppress the interference brought by Gaussian noise.Experimental results show that,on the BOSSbase1.01 dataset,our algorithm significantly outperforms existing mainstream tampering detection algo-rithms in detecting common post-processing operations such as JPEG compression and Gaussian blurring,demon-strating excellent robustness and generalization capabilities.