PENDETEKSIAN DUPLIKASI WILAYAH FOTO DIGITAL MENGGUNAKAN SCALE INVARIANT FEATURE TRANSFORM DAN NEAREST NEIGHBOR
Daftar Isi:
- Copy – move forgery is one of digital photo forgery techniques that is hard to detect by naked eyes, because the forged part came from the same digital photo and blended with the background, especially when rotation and scaling were performed to the forged parts. In this research it is proposed a system to detect fake digital photos, which are original digital photos that has been tempered with copy – move forgeries involving process of translation, rotation or scaling. The input image is a RGB (Red, Green, and Blue) color digital photo that come from CoMoFoD database. At first, the input image is pre-processed into a grayscale image. Then, a feature extraction process is carried out by using Scale Invariant Feature Transform that produce a set of keypoint. After keypoints are found, then the matching process is performed by Nearest Neighbor method, where each keypoint is matched to each other to get the closest distance by using Euclidean Distance and Canberra Distance. The software in this research is developed using the Rational Unified Process method. The experiment and observation results shows that the accuracy of using Nearest Neighbor method with Canberra Distance is higher than the Nearest Neighbor with Euclidean Distance. Overall, fake digital photo with the type of translation is easier to detect than the type of rotation and scalin.