A Computationally Inexpensive Way to Detect and Perform Segmentation on Morphed Images Using CNN

EasyChair Preprint 5569

Abstract

In the last few years, the amount of image data generated by the use of image (and video) processing software such as GNU Gimp, Adobe Photoshop has increased enormously as social networking sites such as Facebook and Instagram have come into force. Such photos are key sources of fake news to propagate radical ideologies, swing public opinion and misused for incitement by the crowd. Being able to solve such a problem and integrating that solution to the social media websites might be able to save the people from becoming a victim of mis-information and propaganda. In this paper we propose a way to detect tampered/morphed images on social media and generate a binary mask of the tampered region using computationally inexpensive pre-processing algorithms and convolutional neural networks

Keyphrases: Classification, Convolutional Neural Network, Fake Images, Segmentation, Tampered Images

@booklet{EasyChair:5569,
  author    = {Neeraj Kumar and Divya Singh and Himja Uppal and Stuti Mehla},
  title     = {A Computationally Inexpensive Way to Detect and Perform Segmentation on Morphed Images Using CNN},
  howpublished = {EasyChair Preprint 5569},
  year      = {EasyChair, 2021}}