无线电工程
無線電工程
무선전공정
RADIO ENGINEERING OF CHINA
2015年
8期
26-29,72
,共5页
曹雁军%李伟%谷宏志%刘晓丽
曹雁軍%李偉%穀宏誌%劉曉麗
조안군%리위%곡굉지%류효려
内容感知%图像缩放%细缝裁剪%图像篡改%马尔科夫特征%支持向量机
內容感知%圖像縮放%細縫裁剪%圖像篡改%馬爾科伕特徵%支持嚮量機
내용감지%도상축방%세봉재전%도상찬개%마이과부특정%지지향량궤
content-aware%image resizing%Seam-carving%image tamper%Markov feature%support vector machine
近几年,内容感知的智能图像缩放技术得到了普遍重视和广泛应用,其中最成熟和流行的细缝裁剪( Seam?Carving)技术可以在尽量保持图像细节不扭曲的情况下进行图像缩放,该技术还可以方便地应用于图像细节的篡改。提出了一种基于马尔科夫特征的细缝裁剪篡改数字图像识别算法,将图像进行8×8分块,对每块进行DCT变换,对所得DCT系数矩阵求取相邻系数水平和竖直方向的差矩阵,利用2个方向的差矩阵得到基于马尔科夫转移概率矩阵的图像特征,并将得到的特征利用支持向量机进行分类训练并对正常图像和细缝裁剪篡改图像进行识别分类。实验结果表明,算法性能优异,可以有效识别正常图像和细缝裁剪篡改图像。
近幾年,內容感知的智能圖像縮放技術得到瞭普遍重視和廣汎應用,其中最成熟和流行的細縫裁剪( Seam?Carving)技術可以在儘量保持圖像細節不扭麯的情況下進行圖像縮放,該技術還可以方便地應用于圖像細節的篡改。提齣瞭一種基于馬爾科伕特徵的細縫裁剪篡改數字圖像識彆算法,將圖像進行8×8分塊,對每塊進行DCT變換,對所得DCT繫數矩陣求取相鄰繫數水平和豎直方嚮的差矩陣,利用2箇方嚮的差矩陣得到基于馬爾科伕轉移概率矩陣的圖像特徵,併將得到的特徵利用支持嚮量機進行分類訓練併對正常圖像和細縫裁剪篡改圖像進行識彆分類。實驗結果錶明,算法性能優異,可以有效識彆正常圖像和細縫裁剪篡改圖像。
근궤년,내용감지적지능도상축방기술득도료보편중시화엄범응용,기중최성숙화류행적세봉재전( Seam?Carving)기술가이재진량보지도상세절불뉴곡적정황하진행도상축방,해기술환가이방편지응용우도상세절적찬개。제출료일충기우마이과부특정적세봉재전찬개수자도상식별산법,장도상진행8×8분괴,대매괴진행DCT변환,대소득DCT계수구진구취상린계수수평화수직방향적차구진,이용2개방향적차구진득도기우마이과부전이개솔구진적도상특정,병장득도적특정이용지지향량궤진행분류훈련병대정상도상화세봉재전찬개도상진행식별분류。실험결과표명,산법성능우이,가이유효식별정상도상화세봉재전찬개도상。
The content?aware image resizing techniques are gaining extensive attention and being widely used these recent years. The Seam?Carving is the most mature one among these techniques,it can resize the image without distorting the materials in image,and also be easily used to tamper the materials in image. To deal with the Seam?Carving tampered image, a detection algorithm based on Markov features is proposed in this paper. The algorithm divides the image into 8 × 8 non?overlapping blocks, performs DCT on every block,gets difference matrix in horizontal and vertical direction from adjacent coefficients in DCT blocks,and then finally obtains fea?tures based on Markov transition probability matrix.These features are trained by SVM and then used to identify the seam?carved images from normal images.The experiment results show that the algorithm is outstanding and can be used to identify Seam?Carved image from normal image effectively.