Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/52791
標題: 植基於支援向量迴歸之影像竄改偵測與還原技術之研究
Image Tamper Proof and Recovery Schemes Based on Support Vector Regression
作者: 蔡垂雄
關鍵字: 資訊科學軟體;應用研究;Image authentication;image tamper detection and recover;fragile watermarking;support vector regression
摘要: 
近年來,數位影像廣泛地應用與傳播於網際網路上,並可視為數位內容呈現之最佳媒介。由於數位影像容易遭受到竄改破壞,造成正確影像內容被竄改後傳遞錯誤內容訊息給接收者。因此,本計畫將針對影像之空間域及頻率域設計一影像竄改偵測與復原系統,用於當影像遭受至竄改後可定位出其竄改區域並進一步將其被竄改區域還原至近似原始影像內容。以下,本計畫將提出兩個影像竄改偵測與還原技術:在第一年計畫中,首先將設計一個以影像區塊與關係為基礎之影像竄改偵測與還原技術,藉由支援向量迴歸學習及預測出來之像素值來嵌入及偵測浮水印位元,並從各像素的LSB位元萃取還原資訊將被竄改影像區塊給復原。在第二年計畫中,則進一步將機制環境設計於小波轉換頻率域中,從小波轉換各子頻帶具有之紋理空間關係作為特徵,以支援向量迴歸學習與預測小波係數並嵌入浮水印位元,同時此預測小波係數也兼具還原資訊之功能。最後,本計畫將研發出結合空間域與頻率域之高強健性、高影像品質之影像竄改偵測與復原技術。

In recent years, digital images have been widely applied and distributedover Internet for a better way to present digital contents. Due to tamper ondigital images could cause incorrect presentations of messages or contents toconfuse or misunderstand receivers, this project proposes an image tamperdetection and recovery system based on the spatial domain and the frequencydomain to achieve above aims. In the first year, an image tamper detection andrecovery based on spatial correlations is proposed. The trained support vectorregression predicts the pixel-value to embed and detect the watermark bits. TheLSB of each pixel is embedded and extracted the recovery information torecover the tampered areas. In the second year, support vector regression isapplied to the wavelet domain to learn the correlations of wavelet sub-bands.Due to the texture correlations between wavelet sub-bands, the waveletcoefficients of sub-bands are used to be the features for support vectorregression to learn and predict wavelet coefficients for watermark embeddingand extraction. Also, the predicted coefficients are used for recovering thetampered areas. At last, this project proposes an image authentication systembased on the spatial domain and the frequency domain for high detecting rateand high recovered image quality.
URI: http://hdl.handle.net/11455/52791
其他識別: NSC99-2221-E005-055-MY2
Appears in Collections:資訊管理學系

Show full item record
 

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.