Dual-Channel Mask-RCNN Model for Image Forgery Detection
With the diversified development of image tampering tools,forged images continue to increase,and are no longer limited to a specific technology such as splicing,copy move,and removal.However,most of the methods currently proposed have poor detection results when they contain multiple types of tampered images.We proposed a dual-channel Mask-RCNN image tamper detection model.The internal statistical features such as noise distribution of the image are extracted through the noise channel,and the surface features such as image contrast differences,tam-pering artifacts and boundaries are captured through the RGB channel.At the same time,the adaptive DA attention module is used to adaptively fuse the features of the two channels to accurately locate the tampered area and achieve pixel-level segmentation.Experimental results on mainstream standard datasets show that the proposed model has bet-ter detection performance than the current advanced model,and is a more general and accurate image forgery detection model.
Image forgery detectionDual-channel networkAttention mechanismNoise information
NETL
NSTL
万方数据
