Image Splicing Detection using HMRF-GMM Based Segmentation Technique and Local Noise Variances

Abstract

Splicing is a well-known, simple and one of the most dangerous images tampering attack. In this attack, an image can be forged by integrating the different parts of the image from the intra orinter images without further post-processing operation such as smoothing of boundaries among different parts of the image. This paper is based on the assumption that images originating from different sources contain different amount of noise, and the camera sensors introduce this noise during image acquisition process. The proposed method estimates local noise variances, and segmentation is performed by Hidden Markov Random Field (HMRF)-Gaussian Mixture Model (GMM). Experimental results demonstrate that proposed method successfully detect splicing on the set of forged images and provide commendable performance.