Detection of image seam carving by using weber local descriptor and local binary patterns

Seam carving is a popular content-aware image resizing technique by removing unnoticeable seams with low energies for aesthetic purpose. However, it might also be used for malicious forgeries such as object removal. In this paper, a blind forensics approach is proposed to detect resized images by seam carving. Since seam carving mainly changes local textures, two excellent texture descriptors including Weber Local Descriptor (WLD) and Local Binary Patterns (LBP) are exploited for seam carving forgery detection. Specifically, the histogram features of WLD and LBP are extracted from candidate images, respectively. Then, Kruskal-Wallis statistic is exploited to select a subset of more discriminative features. Finally, support vector machine (SVM) is exploited as classifier to judge whether an image is original or suffered from seam carving. Extensive experiment results on a large set of test images show that the proposed approach achieves better performance than the state-of-the-art approaches.