Image forensics using EDBTC feature

This paper presents a new method on image forensics application using the Error Diffusion Block Truncation Coding (EDBTC) feature. The image forensics tries to detect the copy-move forgery image regions on the forged image. Firstly, an image is divided into several non-overlapping image blocks. The image feature is further derived for each image block. Herein, two image features, namely Color Feature (CF) and Bit Feature (BF), are composed from the EDBTC compressed data stream. The forged region is detected while the image feature of this region is similar to the image feature of other region separated far away. As documented in the experimental section, the proposed method gives a promising result on the image forensics taks, and, at the same time, outperforms the former existing scheme.