Detection of duplicated regions in digital images has been a highly investigated field in recent years since the editing of digital images has been notably simplified by the development of advanced image processing tools. In this paper, we present a new method that combines Cellular Automata (CA) and Local Binary Patterns (LBP) to extract feature vectors for the purpose of detection of duplicated regions. The combination of CA and LBP allows a simple and reduced description of texture in the form of CA rules that represents local changes in pixel luminance values. The importance of CA lies in the fact that a very simple set of rules can be used to describe complex textures, while LBP, applied locally, allows efficient binary representation. CA rules are formed on a circular neighborhood, resulting in insensitivity to rotation of duplicated regions. Additionally, a new search method is applied to select the nearest neighbors and determine duplicated blocks. In comparison with similar methods, the proposed method showed good performance in the case of plain/multiple copy-move forgeries and rotation/scaling of duplicated regions, as well as robustness to post-processing methods such as blurring, addition of noise and JPEG compression. An important advantage of the proposed method is its low computational complexity and simplicity of its feature vector representation.
展开▼
