Image tampering detection method and device based on semantic irrelevant feature learning

• Main Authors: TANG SHENG, CAO JUAN, SUN ZHIHAO, WANG DANDING
• Format: Patent
• Language: Chinese; English
• Published: 21.11.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The invention provides an image tampering detection method and device based on semantic irrelevant feature learning, and the method comprises the steps: extracting the semantic features of a to-be-detected image through a pre-trained reference semantic feature encoder, and constraining the similarity between a local region tampering trace feature and a reference semantic feature through a comparative learning framework, the semantic correlation in the tampering trace features is directly limited; the model is guided to mine the inconsistency between the real area near the tampered area boundary and the tampered area feature by means of tampered area boundary supervision, and a feature conversion structure is designed to realize the unification of the auxiliary task and the main task so as to ensure that the tampered area boundary prediction task serving as the auxiliary task can provide gain for the tampered area prediction task. 本发明提出一种基于语义无关特征学习的图像篡改检测方法和装置，包括通过预训练的基准语义特征编码器提取待检测图像的语义特征，使用对比学习框架约束局部区域篡改痕迹特征