A passive image forensic scheme based on an adaptive and hybrid techniques
Manaf Mohammed Ali Alhaidery,Amir Hossein Taherinia
DOI: https://doi.org/10.1007/s11042-022-12374-5
IF: 2.577
2022-02-21
Multimedia Tools and Applications
Abstract:The fast growth of using digital images in the virtual world and, The advance of image processing tools makes an image under the effect of attackers or intruders. Attackers tamper digital images and divert the content of the image from its true meaning. The manipulation in digital images makes tampering hard to detect with the naked eye. Therefore, image forensic analysis is developed to keep up and protect the authenticity and rights of the owner of digital images. The Copy-move forgery detection (CMFD) scheme is the popular type in image forensic analysis. This paper presents a merged CMFD scheme. This scheme contains a preparing process and three layers of processing. In the preparing process, Haar Discrete Wavelet Transform(HDWT) automates the number of segments before applying the segmentation process. The first layer, Simple Linear Iterative Clustering(SLIC)segmentation method is proposed to split images into irregular labels. Then, classification these segments into two main types called texture or smooth regions based on the Entropy metric. In the second layer, the keypoint-based technique is adopted. Speed Up Robust Feature(SURF) as detector and Histogram Of Oriented Gradient (HOG) as a descriptor is applied on each region. SURF-HOG to extract key points from regions with different Threshold values. In the third layer, an efficient probabilistic false positive removal filter(M-SAC) is employed. It aims to include the correct results and exclude false results from the output detected image. Subsequently, increase TPR and decrease FPR. The proposed scheme is evaluated by using(IMD and MICC-F220)data sets. The photometric attacks(brightness, blurring, JPEG compression) and geometric transformation attacks(scaling, rotation) are applied in these datasets. The experimental results indicate that the proposed scheme is fast, efficient, and high performance under simple and compound attacks. It has high TPR, Low FPR and, makes the scheme more dynamic and suitable in image forensic analysis.
computer science, information systems, theory & methods,engineering, electrical & electronic, software engineering