基於多重分類及區塊擴張之複製移動竄改偵測

機構典藏 > College of Engineering > Graduate Institute & Department of Computer Science and Information Engineering > Thesis > Item 987654321/105742

Please use this identifier to cite or link to this item: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/105742

Title:	基於多重分類及區塊擴張之複製移動竄改偵測
Other Titles:	An efficiency enhanced cluster expanding block algorithm for copy-move forgery detection
Authors:	張逸丞;Chang, Yi-Cheng
Contributors:	淡江大學資訊工程學系碩士班陳建彰;Chen, Chien-Chang
Keywords:	複製移動竄改偵測;多重分類;區塊擴張;Copy-move forgery;Enhanced Cluster;Expanding block
Date:	2015
Issue Date:	2016-01-22 15:04:03 (UTC+8)
Abstract:	本論文提出一個以多重分類為基準的複製移動竄改偵測技術，在多重分類的方法流程中，首先對影像所分割出來的區塊進行兩種不同的特徵提取，接著以兩種特徵值對區塊分別進行分類，並且將每個類別集合成大類以增加區塊相似度的範圍，透過多重分類篩選相似的區塊後，取在兩種分類中皆為同類別的區塊進行比對，可以有效的降低區塊間比對次數，使得在比對階段效率大幅提升，且同時過濾不相似區塊。比對階段時，利用由小到大的區塊擴張比對方式，可以在大量的比對動作中，有效減少其計算量，最後將偵測出區塊以記綠水平垂直距離的方式進行匹配，以找出被竄改區域。本論文在不降低正確率的前提下，提升了整體執行效率。 This paper presents an algorithm for identifying copy-move forgery area in an image using efﬁcient enhanced cluster expanding blocks strategy. First, the image is divided into overlapping blocks and two features, mean and variance, are calculated from each block. These two features are sorted and grouped into two different kinds of clusters. Combining a cluster with its neighboring cluster forms a bucket. The proposed scheme then uses expanding block method to compare all blocks in each bucket. Finally, the duplicated regions are found by matching detected blocks horizontally and vertically. Experimental results show that the proposed approach detects duplicated regions well under JPEG compression and Gaussian blurring.
Appears in Collections:	[Graduate Institute & Department of Computer Science and Information Engineering] Thesis

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	113	View/Open

Loading...