Image Forgery Detection Using Back Propagation Neural Network Model and Particle Swarm Optimization Algorithm
Merlin Gladiss, O Mayer, M Stamm, Y Li, J Zhou, B Chen, M Yu, Q Su, H Shim, Y Shi, K Hosny, H Hamza, N Lashin, J Bappy, C Simons, L Nataraj, B Manjunath, A Roy-Chowdhury, S Jia, Z Xu, H Wang, C Feng, T Wang, Bin Xiao, Yang Wei, Xiuli Bi, Weisheng Li, Jianfeng Ma, Victor Schetinger, Massimo Iuliani, Alessandro Piva, M Manuel, Oliveira, M Stamm, M Wu, K Liu, A Popescu, H Farid, H Farid, M Stamm, K Liu, M Kirchner, J Fridrich, X Kang, M Stamm, A Peng, K Liu, M Chen, J Fridrich, J Lukas, M Goljan, J Fridrich, D Soukal, J Lukas, I Amerini, L Ballan, R Caldelli, A Del, G Bimbo, Serra, I Amerini, L Ballan, R Caldelli, A Del, G Bimbo, Serra, Tiago Jos De Carvalho, Christian Riess, Elli Angelopoulou, Hlio Pedrini, Anderson De, Rezende Rocha, P Felzenszwalb, D Huttenlocher, G Buchsbaum, Mridul Kumar, Mathur, Priyanka Bhati, Renjith Thomas, M Rangachar, L Yan, P Hu, C Li, Y Yao, L Xing, F Lei, Xiwen Cai, Liang Gao, Fan Li
Abstract
From the images, forgery detection is ahead noteworthy attention as there are numerous editing tools that facilitate to form the edition by means of removal or objects manipulation from the images. These editing tools not at all abscond any forgery trace as a result, growing the challenges of the system to identify the existence of the manipulations. Accordingly, this work presents a new forgery recognition approach which is on the basis of the supervised learning algorithm. This algorithm is presented concerning exploiting the BPNN and optimization is set up by exploiting the Particle Swarm Optimization (PSO) algorithm. The classification is done by exploiting the proposed classifier exploits the texture features attained from the GWTM descriptor so that the features are attained from the face-detected images retrieved by exploiting the Viola-Jones approach. By utilizing two datasets, the simulation is performed to show the efficiency of the proposed model. By utilizing the datasets the analysis shows that the proposed model obtained superior performance while evaluating with few conventional forgery detection systems.