结合游程长度与共生矩阵的图像拼接篡改检测方法

摘要

An algorithm on blind detection of splicing images by using the statistical characteristics with run-length and gray level co-occurrence matrix(GLCM) features was researched.Splicing detection can be treated as a two-class pattern problem.This paper proposed an enhanced technique for blind detection of image splicing.It extracts the run-length run-number(RLRN) in CbCr chroma space and combines with GLCM feature to detect the artifacts introduced by the tampering operation.Then,the LibSVM is exploited to classify the authentic and spliced images.The experiment results demonstrate that the proposed algorithm not only improves the accurate recognition rate in CASIA v2.0 dataset,but also can achieve well copy-move tampered detection.%针对拼接图像篡改鉴别问题,研究了运用游程长度和灰度共生矩阵等统计特征进行图像盲检测的算法.该算法将拼接篡改检测问题转换为二分类模式识别问题,在图像的色度通道Cb和Cr中提取游程长度特征,并将其与灰度共生矩阵特征相结合,然后以结合后的新的特征向量训练LibSVM作为最后的分类器,确定所给图像是否被拼接篡改.实验结果显示,在彩色图像测试库CASIA v2.0上,利用融合后的特征检测图像使识别率得到较大提高,同时该算法在复制粘贴篡改方面也有较好的识别性能.