الفهرس | Only 14 pages are availabe for public view |
Abstract an efficient blind detection algorithm is proposed for the detection of forgeries. This algorithm is based on Homomorphic image processing which transforms the image into illumination and reflectance component. The illumination component is constant or has small variation depending on whether the image is authentic or tampered. SVM and ANN classifiers are used for this process and a comparison study between them are investigated. ROC is used to evaluate classifiers performances. Three different image sets are examined. The first consists of digital images acquired from digital cameras. The second set consists of digital images scanned from scanners. The final set consists of the two image types. Three different color coordinate systems are tested using the proposed algorithm and their results are compared to obtain the highest accuracy level. The performance of the proposed algorithm is tested with JPEG compressed images and also in presence of noise. A comparison between the proposed approach (algorithm) with four previously published ones shows that proposed is more general one and gives high detection accuracy regardless the source of acquisition device. |