本文關(guān)鍵詞： 區(qū)域復(fù)制粘貼篡改 塊匹配 Hu矩 Zernike矩 特征點(diǎn)匹配 SIFT算法 高斯幾何不變矩　出處：《山東大學(xué)》2017年碩士論文　論文類型：學(xué)位論文

【摘要】：隨著數(shù)碼相機(jī)、高性能智能手機(jī)的普及和高性能攝像頭的不斷革新,數(shù)字圖像作為日常生活中的信息載體越來越得到普遍應(yīng)用。隨著數(shù)字圖像處理編輯軟件也大量出現(xiàn),雖然對(duì)圖像處理帶來了很大的益處,但任何事情都是"雙刃劍",這也給那些心懷叵測之人帶來了危害社會(huì)的更加容易的手段。目前,網(wǎng)絡(luò)上和社會(huì)上充滿著大量偽造圖片在混淆視聽,已經(jīng)對(duì)人們的切身利益造成了傷害。是以,針對(duì)偽造圖像的識(shí)別與研究具備重大價(jià)值,也應(yīng)該提到探索日程上來。當(dāng)前圖像篡改檢測算法琳瑯滿目,我國關(guān)于這方面的研究雖然取得了非常大的成就,但目前還處于初步階段,也存在許多不足之處,算法性能比較單一。本文中主要針對(duì)數(shù)字圖像區(qū)域復(fù)制粘貼篡改手段盲檢測研究,區(qū)域復(fù)制粘貼篡改屬于局部篡改手段類別。目前,針對(duì)圖像區(qū)域復(fù)制粘貼篡改盲檢測技術(shù)主要分為基于圖像塊檢測算法和基于特征點(diǎn)檢測算法,本文對(duì)前人算法不足進(jìn)行分析,并提出相應(yīng)的改進(jìn)算法。論文主要工作:1、論文首先具體講述了關(guān)于數(shù)字圖像篡改認(rèn)證技術(shù)研究背景與意義、國內(nèi)外研究現(xiàn)狀,系統(tǒng)的分析了圖像篡改的方式,著重介紹了同幅圖像copy-move篡改手段,詳細(xì)分析了圖像copy-move篡改模型以及概述現(xiàn)存對(duì)copy-move篡改盲取證技術(shù)認(rèn)證算法,詳細(xì)介紹了現(xiàn)存較為經(jīng)典檢測方法。2、本文第三章針對(duì)圖像盲檢測技術(shù)基于圖像塊檢測算法,分析現(xiàn)存圖像塊算法的優(yōu)缺點(diǎn)基礎(chǔ)上,提出了基于改進(jìn)Hu矩和Zernike矩結(jié)合的圖像塊匹配法。該算法基于改進(jìn)Hu矩和Zernke矩來表征特征向量,該算法首先對(duì)圖像進(jìn)行滑動(dòng)分塊,提取圖像塊特征向量,利用圖像子塊特征向量相關(guān)性來識(shí)別圖像篡改且定位其位置,該算法實(shí)時(shí)性得到提高,且對(duì)平移、旋轉(zhuǎn)魯棒性較好。實(shí)驗(yàn)結(jié)果表明該算法可以有效抵制對(duì)篡改區(qū)域進(jìn)行旋轉(zhuǎn)、平移操作處理。3、本文第四章針對(duì)圖像盲檢測算法基于特征點(diǎn)檢測算法,分析傳統(tǒng)SJIFT算法缺陷,提出了融合高斯幾何不變矩的改進(jìn)SIFT的特征點(diǎn)圖像篡改取證算法。該算法中先用改進(jìn)SIFT算法提取圖像關(guān)鍵點(diǎn),為特征點(diǎn)分配主方向,然后提取關(guān)鍵點(diǎn)鄰域窗口的高斯幾何不變矩作為關(guān)鍵點(diǎn)的特征描述子,最后進(jìn)行特征描述子的匹配,該算法采用歐氏距離進(jìn)行特征點(diǎn)匹配,并利用自適應(yīng)歐氏距離閾值與RANSAC結(jié)合算法剔除誤匹配對(duì),實(shí)現(xiàn)篡改區(qū)域的識(shí)別與定位。實(shí)驗(yàn)結(jié)果表明該算法可以基本保持圖像提取的特征點(diǎn)數(shù),甚至?xí)䴗p少提取的特征點(diǎn)數(shù),但可以增加特征點(diǎn)匹配點(diǎn)數(shù),減少誤匹配點(diǎn),并且可以提高特征點(diǎn)提取時(shí)間,因?yàn)樘卣髅枋鲎泳S數(shù)的減少,對(duì)于匹配效率也有一定的提高,實(shí)驗(yàn)效果圖可以看出,該算法對(duì)于復(fù)制區(qū)域的平移、尺度縮放與旋轉(zhuǎn)操作都具有非常好的魯棒性,檢測精度也較高。
[Abstract]:With the popularization of digital camera, high-performance smart phone and the innovation of high performance camera, digital image is becoming more and more popular as information carrier in daily life. While there are great benefits to image processing, everything is a "double-edged sword," which also gives those who have evil intentions an easier means of harming society. There are a lot of fake images on the Internet and in society, which are confusing and harmful to people's vital interests. Therefore, it is of great value to identify and study fake images. We should also mention the agenda of exploration. At present, there are a great variety of algorithms for image tampering detection. Although great achievements have been made in this area in our country, it is still in its preliminary stage, and there are also many deficiencies. The performance of the algorithm is relatively simple. This paper mainly focuses on the blind detection of digital image region copy-paste tamper, which belongs to the category of local tamper. The blind detection technology of image region copy and paste tamper is mainly divided into image block detection algorithm and feature point detection algorithm. In this paper, the shortcomings of previous algorithms are analyzed. The thesis mainly focuses on the research background and significance of digital image tampering authentication technology, the current research situation at home and abroad, and the systematic analysis of image tampering methods. This paper mainly introduces the copy-move tampering method of the same image, analyzes the image copy-move tamper model in detail, and summarizes the existing authentication algorithms for copy-move tampering blind forensics technology. This paper introduces the existing classical detection method. 2. The third chapter analyzes the advantages and disadvantages of the existing image block detection algorithm based on the image block detection algorithm. An image block matching method based on improved Hu moments and Zernike moments is proposed, which is based on improved Hu moments and Zernke moments to represent feature vectors. By using image subblock feature vector correlation to identify image tampering and locate its position, the real-time performance of the algorithm is improved, and the robustness to translation and rotation is good. Experimental results show that the algorithm can effectively resist the rotation of tampered regions. Translation operation processing. 3. In Chapter 4th, aiming at blind image detection algorithm based on feature point detection algorithm, the defects of traditional SJIFT algorithm are analyzed. An improved feature point image tampering and forensics algorithm based on Gao Si's geometric invariant moment is proposed in this paper. The improved SIFT algorithm is first used to extract the key points of the image and assign the main direction to the feature points. Then the Gao Si geometric invariant moment of the neighborhood window of the key points is extracted as the feature descriptor of the key points. Finally, the feature descriptors are matched, and the Euclidean distance is used to match the feature points. The adaptive Euclidean distance threshold and the RANSAC algorithm are used to eliminate the mismatch pairs to realize the tamper recognition and localization. The experimental results show that the proposed algorithm can basically keep the feature points extracted from the image and even reduce the extracted feature points. But it can increase the matching points of feature points, reduce the mismatch points, and improve the extraction time of feature points, because the reduction of subdimension of feature description can also improve the matching efficiency. The algorithm is robust to the translation, scaling and rotation operations of the replication region, and the detection accuracy is also high.
【學(xué)位授予單位】：山東大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：TP391.41

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 陳勇;趙惠昌;陳思;張淑寧;;基于SIFT彈載SAR圖像匹配算法[J];系統(tǒng)工程與電子技術(shù);2016年06期

2 陳竺益;方針;;基于色像差特性的圖像篡改檢測[J];應(yīng)用科學(xué)學(xué)報(bào);2015年06期

3 趙潔;郭繼昌;;基于JPEG系數(shù)變化率的圖像復(fù)制粘貼篡改檢測[J];浙江大學(xué)學(xué)報(bào)(工學(xué)版);2015年10期

4 李巖;劉念;張斌;袁開國;楊義先;;一種基于逆序廣義2近鄰的圖像多重復(fù)制粘貼篡改檢測算法[J];電子與信息學(xué)報(bào);2015年07期

5 李巖;劉念;張斌;袁開國;楊義先;;圖像鏡像復(fù)制粘貼篡改檢測中的FI-SURF算法[J];通信學(xué)報(bào);2015年05期

6 張曉琳;方針;張新鵬;;利用通道間相關(guān)性的CFA圖像盲取證[J];應(yīng)用科學(xué)學(xué)報(bào);2015年01期

7 Ying Li;Bo Wang;Xiang-Wei Kong;Yan-Qing Guo;;Image Tampering Detection Using No-Reference Image Quality Metrics[J];Journal of Harbin Institute of Technology;2014年06期

8 王青;張榮;;基于DCT系數(shù)雙量化映射關(guān)系的圖像盲取證算法[J];電子與信息學(xué)報(bào);2014年09期

9 張朝鑫;席平;;高斯幾何矩及其在特征匹配與圖像配準(zhǔn)中的應(yīng)用[J];計(jì)算機(jī)輔助設(shè)計(jì)與圖形學(xué)學(xué)報(bào);2014年07期

10 鞏道福;劉粉林;羅向陽;;一種基于圖像邊緣的魯棒水印算法[J];中國科學(xué):信息科學(xué);2013年11期

相關(guān)博士學(xué)位論文 前1條

1 呂穎達(dá);數(shù)字圖像盲鑒別的關(guān)鍵理論與技術(shù)研究[D];吉林大學(xué);2015年

相關(guān)碩士學(xué)位論文 前3條

1 孫愛華;數(shù)字圖像來源與篡改檢測算法研究[D];中國海洋大學(xué);2014年

2 單薇;基于復(fù)制粘貼的數(shù)字圖像篡改檢測研究[D];蘇州大學(xué);2014年

3 劉迪;基于流形學(xué)習(xí)與子空間的降維方法研究與應(yīng)用[D];東北師范大學(xué);2009年

，

本文編號(hào)：1509017