基于LPP和改进SIFT的copy-move篡改检测

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。 

Copy-move Forgery Detection Based on LPP and Improved SIFT Algorithm

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.