基于LPP和改进SIFT的copy-move篡改检测

摘要

图像的复制-粘贴篡改检测是图像篡改检测领域中的重要组成部分.本文基于SIFT算法以及LPP的降维思想,提出了一种新的篡改检测算法.本文在SIFT算法的基础上,使用LPP算法对SIFT算法生成的特征点以及特征向量进行降维.使得传统SIFT算法在实际应用中特征点数目过多、特征向量维数过高等缺陷得到了解决.并使用凝聚型层次聚类算法对相似的特征点进行聚类,完成了对图像复制-粘贴篡改区域的检测.在文章的最后,本文对哥伦比亚大学复制-粘贴图像库里的100张图片进行实验.实验结果表明,不管篡改区域后处理方式是拉伸还是旋转,本文算法都能比传统的SIFT、SURF、PCA-SIFT等算法生成更少的特征点数目和更低的特征向量维度,使得检测效率以及检测正确率得到有效提升.%Image copy-move forgery detection is an important part of image forgery detection.In this paper,we propose a novel forgery detection algorithm based on the SIFT and LPP algorithm.First,we use LPP algorithm to reduce the dimension of feature points and feature-vectors which are generated by the SIFT algorithm.Solving many defects of the traditional SIFT algorithm like the number of feature points are too many,the dimension of the feature-vector is too high etc.Then we cluster the similar feature points using the cohesive hierarchical clustering algorithm to find out the copy-move forgery area.At the end of the article,we make some experiments to test our algorithm,using 100 pictures in the Columbia University copy-move forgery image library.Results show that the proposed algorithm can generate fewer feature points and lower dimensions of the feature-vector than traditional SIFT,SURF,and PCA-SIFT algorithm,making the efficiency and the accuracy rate of the image forgery detection greatly improved,regardless of how the forgery regions are stretched or rotated.