CAJ | 학술논문

针对比较常见的图像的复制-粘贴篡改技术，提出一种基于局部二值模式LBP（ local binary pattern ）的检测算法。首先把需要检测的已经被篡改的图像分成大小相同的重叠块，每块的纹理特征用LBP（旋转不变）向量去表示，从而得到被检测图像的特征矢量；然后对得到的特征矢量进行字典排序，并结合检测图像块的位移矢量，准确定位并检测出图像中的被篡改区域。实验结果表明：在抗旋转处理和效率方面该算法均优于经典的基于PCA的检测算法。
침대비교상견적도상적복제-점첩찬개기술，제출일충기우국부이치모식LBP（ local binary pattern ）적검측산법。수선파수요검측적이경피찬개적도상분성대소상동적중첩괴，매괴적문리특정용LBP（선전불변）향량거표시，종이득도피검측도상적특정시량；연후대득도적특정시량진행자전배서，병결합검측도상괴적위이시량，준학정위병검측출도상중적피찬개구역。실험결과표명：재항선전처리화효솔방면해산법균우우경전적기우PCA적검측산법。
Aiming at quite common technique of image copy-move forgery, in this paper we propose a detection algorithm which is based on local binary pattern .First, we divide the tampered image to be detected into multiple overlapping blocks with the same size , the textural features on each block are represented with rotation invariant LBP vectors , therefore the feature vectors of the detecting image are got . Secondly , we sort the derived feature vectors in dictionary order , locate and detect the tampered region in the image by combining the displacement vectors of image blocks .Experimental results show that our algorithm performs better than the classical detection algorithm based on PCA ( principal component analysis ) in terms of the robustness against rotating operation and the efficiency .