Abstract
Over the last few years, manipulation of digital images for forgery purpose with high quality has become popular using low-cost acquisition devices. Discrimination between tampered images and authentic images is required to ensure the origin of the data carried by images. In this paper, high-pass filtering and histogram equalization are used as pre-processing stages to reinforce the details of the images prior to forgery detection. The detail reinforcement process can achieve high classification accuracy. Illumination histogram is estimated after pre-processing and the peak value of the histogram derivative is estimated and used as a metric for forgery detection. The philosophy behind this strategy is that in the case of forgery, the forged regions are subject to different illumination conditions, and hence this can be reflected in the illumination histogram peaks. A thresholding approach is adopted in this paper. Probability Density Functions (PDFs) of the histogram derivative peaks are adopted for classification.