Copy–Move Image Multiple Forgery Detection Based on Transit Flow Regime Algorithm–Enabled ShuffleNet.

The copy-move forgery is considered one among difficult kinds of image forgeries that must be detected. This occurs by duplicating image parts or portions and thereafter adding up again in an image by itself but in other locations. The forgery detection techniques are utilized in image protection while an actual image is not obtainable. The current forgery detection methods detect tampered areas with lesser effectiveness owing to larger size as well as low contrast of images. Here, transit flow regime algorithm-based ShuffleNet (TFRA-ShuffleNet) is presented for multiple forgery detection. In this work, the input image considered is given to multiple object detection. The multiple object detection is carried out in an input image utilizing YOLO V3. After that, features are extracted from object detected image that include local vector pattern (LVP), local optimal-oriented pattern (LOOP), pyramid histogram of oriented gradients (PHoG), local Gabor XOR patterns (LGXP), local directional pattern (LDP), local directional ternary pattern (LDTP) and local binary pattern (LBP). Lastly, multiple forgery detection is conducted employing ShuffleNet. The ShuffleNet is trained to employ TFRA, which is an integration of transit search (TS) and flow regime algorithm (FRA). Furthermore, TFRA-ShuffleNet achieved maximal accuracy, TPR and TNR values of about 96.5%, 96.5% and 97.5%. [ABSTRACT FROM AUTHOR]