基于模糊局部二值模式算子的图像伪造检测

摘要

To solve these problems that detection precise is influenced by the threshold parameter,and the image feature mis-match can not be eliminated,as well as robustness is poor,the forgery detection algorithm based on random sample consensus strategy coupled fuzzy local binary pattern operator was proposed.The per-pixel adaptive filter was introduced to filter image blocks.The fuzzy local binary pattern operator was designed to extract the image features by introducing the fuzzy set theory. The lexicographical sorting was embedded to sort these feature vectors.The effective block matching algorithm was designed to accurately detect the real forgery block by embedding the K-d tree and defining the majority rules.The random sample consensus strategy was used to eliminate mismatching of image blocks.Simulation results show that compared with current forgery detec-tion technology,this algorithm has higher detection accuracy and stronger robust.%为解决当前图像复制移动伪造检测算法的识别精度受阈值参数影响较大、无法消除图像特征误配、以及鲁棒性不佳等不足，提出随机样本一致策略耦合模糊局部二值模式算子的图像伪造检测算法。引入逐像素自适应 Wiener 滤波，过滤图像分块；基于局部二值模式 LBP （local binary patterns），引入模糊集理论，设计模糊二值模式算子，提取图像特征；基于字典排序，对特征矢量进行排序；嵌入 K-d 树，定义识别多数规则，设计高效块匹配算法，精确检测真实的伪造分块；利用随机样本一致算法，消除图像分块误配。仿真结果表明，与当前伪造检测技术相比，该算法的检测精度更高，具有更强的鲁棒性。