重庆科技大学学报（自然科学版）2026，Vol.28Issue(1)：70-79,10.DOI:10.19406/j.issn.2097-4531.2026.01.008
基于PVM的Seam Carving篡改检测算法
PVM-Based Seam Carving Tamper Detection Algorithm
摘要
Abstract
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.关键词
接缝裁剪/后处理场景/图像篡改检测/Vision Mamba/鲁棒性Key words
Seam Carving/post-processing scenarios/image tampering detection/Vision Mamba/robustness分类
信息技术与安全科学引用本文复制引用
赵洁,常皓婵,王浩天,武斌..基于PVM的Seam Carving篡改检测算法[J].重庆科技大学学报（自然科学版）,2026,28(1):70-79,10.基金项目
天津市科技计划项目"大数据环境下智慧城市视频监控智能分析平台关键技术研究"(19JCTPJC47200) （19JCTPJC47200）
