Digital Image Forensics Using Local Optimal-Oriented Pattern and ELM

Abstract

Due to the advances in computer graphic technology, it has become very difficult to classify computer graphics from photographic images. It is also very difficult to distinguish tampered images from authentic images. In the chapter, an algorithm has been proposed to solve both the problems simultaneously. A texture descriptor local optimal-oriented pattern has been used to extract features from computer-generated and photographic images. The extreme learning machine has been used to classify the images. Experiment results demonstrate that it has high accuracy and is also robust against post-processing operations.