结合关键点和块优点的复制粘贴检测算法  

Copy-move Detection Algorithm Combining Keypoints and Block Advantages

作　　者：杨珊珊 张大兴[1] 郭家伟 王诗迢 YANG Shanshan;ZHANG Daxing;GUO Jiawei;WANG Shitiao(Institute of Image Graphics,Hangzhou Dianzi University,Hangzhou 310018,China)

机构地区：[1]杭州电子科技大学图形图像研究所

出　　处：《电子科技》2020年第3期38-43,共6页Electronic Science and Technology

基　　金：国家自然科学基金(61572160)~~

摘　　要：数字图像应用广泛,但随着各种编辑图像软件的使用,使得图像的真实性有待考证。大多数基于关键点检测的算法由于关键点数量有限,导致最后检测结果不完整或者表示不明确,文中提出了一种基于分割结合关键点特征的检测算法,利用SIFT提取关键点特征,再使用g2NN算法匹配关键点,用设计的聚类去除误匹配,根据改进的自适应分割算法,标记匹配的图像块,然后找出周围邻居块进行对比匹配输出结果。该算法能有效快速地检测并标记大图片的复制篡改区域,对大多数后处理鲁棒,较完整地标记出篡改区域。Digital images are widely used,and the use of various software for editing images has led to the authenticity of the images to be verified.Most of the algorithms based on keypoint detection have a limited number of key points,resulting in incomplete or unclear final detection results.This paper proposed a detection algorithm based on segmentation combined with keypoint features.First,SIFT was used to extract keypoint features and then used.The g2NN algorithm matched the key points,used the designed clustering to remove the mismatch,and according to the improved adaptive segmentation algorithm,marked the matched image blocks,and then finded the surrounding neighbor blocks to compare and matched the output results.The algorithm could effectively detected and marked the copy tampering area of large pictures,and was robust to most post-processing,and marked the tampering area more completely.

关 键 词：复制粘贴检测 SIFT 分割 关键点匹配 聚类 鲁棒 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]