Automated Image Splicing Detection using Texture based Feature Criterion and Fuzzy Support Vector Machine based Classifier

摘要

Nowadays manipulation in digital image becomes quite simple and convenient with the broad use of advanced image editing software and the recognition of forged region cannot be captured by naked human eyes easily. One of the most frequent alterations apply on the images is called image splicing, which consists of cropping a region from an image and paste it onto a different image in order to alter the information present in the new source. There are various methodologies which address this kind of attack, by dividing the image into blocks in an attempt to detect the inconsistencies created by the attack. Though features can be edges, corners, blobs or characteristics of noise changes, alteration in illumination acquired throughout the image. Moreover image splicing is quite harder to identify tampering in a forged image. In this paper the methodology used for splicing detection is relied on texture based features which are extracted from the tampered and actual images in the form of Grey Level Run Length Matrix calculated in four directions. The features have demonstrated the potential of proposed methodology used for identification of image forgery. Thereafter fuzzy support vector machine is utilized as a new classification technique with the high generalization capability. More effective methodology is proposed for detection of splicing forgery and produced improved classification outputs.