Open AccessCMFD using a Novel Localisation Technique and CNN based Classification
Author(s) -
Ritu Agarwal
Publication year - 2019
Publication title -
international journal of engineering and advanced technology
Language(s) - English
Resource type - Journals
ISSN - 2249-8958
DOI - 10.35940/ijeat.a2019.109119
Subject(s) - artificial intelligence , block (permutation group theory) , computer science , pattern recognition (psychology) , convolutional neural network , scale invariant feature transform , image (mathematics) , software , contextual image classification , computer vision , gaussian filter , mathematics , geometry , programming language
Latest trends of the image processing software, the growth of image manipulation is at peak. To detect the use of such software on an image is a growing research anomaly. This paper proposes a novel copy-move forgery localization approach in an image through a blind approach with no prior information available to the algorithm. Here, we have split the image into equal size blocks and extracted SIFT features for every block. The center of mass for each block is calculated after applying the Gaussian filter. Finally, image features are matched based on the KNN algorithm for CMF localization. However, for classification, the localisation mask is created for the dataset, and is used to train a Convolutional neural networks(CNN) and this trained CNN in turn is used for classification of images as authentic or tampered.