Detecting Digital Image Origin: Photographic or Photorealistic

Abstract

With the rapid technological evolution that has taken place in the last decade, a genuine question has arised-whether the information being propagated is a true account or doctored. Digital images are being generated at a tremendous pace and thus face the issue of their authenticity. The application of image forensics ranges from fields like legal enforcement , entertainment industry, and politics, among many others.

Over the years, extensive research has been done to detect the origin of digital images - camera or CGI . Further operations such as detecting the camera make and even the specific model that was used for capturing the image are also being investigated. Another aspect of image forensics lies in detecting whether a camera image is authentic or has been tampered with.

Keeping the above objectives in mind, two techniques were implemented. Both work on the basic principle of detecting demosaicing traces in an image , which is proof of it being a camera image. One utilizes Fourier Analysis , and the other technique utlizes Adaptive Delta Local Rank Transform . A SVM classifier was used to test the validity of the above methods.

The methods were computationally simpler than most techniques in use today , and provided a high degree of accuracy.