Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/72818
標題: 可預測之可回復式資訊隱藏演算法
A+Predictable+Scheme+for+Reversible+Data+Hiding
作者: 黃敬儒
Huang, Jing-Ru
林昀德
王宗銘 
Lin, Yun-Te
Wang, Chung-Ming
關鍵字: 預測機制;prediction mechanism;可回復式資訊隱藏;資訊隱藏;偽裝學;reversible data hiding;data hiding;steganography
出版社: 國立中興大學工學院;Airiti Press Inc.
Project: 興大工程學刊, Volume 23, Issue 3, Page(s) 119-144.
摘要: 
現存預估器差值擴張法存有三種問題亟待解決。首先,我們無法預先得知透過嵌入程序後的結果。其二為當使用者有特殊需求,無法快速取得相關參數並執行嵌入程序,唯一辦法僅能使用錯誤嘗試方式來循序改變參數、執行嵌入並驗證是否滿足需求。第三為無法快速從多個掩護影像中找到一個具有最佳嵌入效果或最適合使用者需求的掩護影像。本文提出一個預測機制,改善上述缺點。我們預測機制先建構五個直方圖,並利用直方圖來分析並建立期望均方誤差與期望淨嵌入量的數學函式。我們也將秘密訊息"0"或"1"出現機率納入考量。據此,我們的預測機制可以自動預測單張掩護影像之最佳嵌入參數與最佳嵌入結果。此外,當使用者有特殊需求時,僅需輸入特定的參數,我們的機制經過簡單的數學計算,即可快速產出對應之參數,進一步預測得出符合使用者需求之嵌入結果。最後,我們將機制延伸至影像資料庫,從中自動選出具有最佳嵌入結果之掩護影像。該機制也提供使用者從多張掩護影像中選出最適合其需求之掩護影像並產出預測之嵌入結果。使用14張測試影像之實驗結果顯示:我們的預測機制具有極高的準確性,扣除嵌入成本後之淨嵌入量與實際嵌入秘密訊息所得之淨嵌入量,其平均誤差率小於10^(-6)%;預測的影像品質也具有極高的正確率,預測與實際得出之峰值信噪比,其平均誤差率小於10^(-6)%。對於資料庫中之影像,也可以精準的排序,預測能提供最大嵌入量之掩護影像。總結本文,我們提出的預測機制具有預測性、便利性、最佳性、廣泛性與整合性等五個特徵,具體可行,擴大可回復式資訊隱藏之應用範疇。

Existing reversible data hiding algorithms using the prediction error approach have three unsolved problems. In this paper, we introduce a prediction mechanism to solve these problems. Our prediction mechanism analyzes a given cover image and constructs five histograms first before we derive mathematical functions to express the expected mean square error (EMSE) and expected pure capacity (EPC). Our prediction scheme takes into consideration the appearance of the probability of the secret message "1." Consequently, the mechanism can predict the optimal data hiding results prior to the real secret message embedding. In addition, the scheme can satisfy a user's demand for pure capacity or stego image quality. In particular, the derived mathematical functions are computed so as to automatically suggest the most appropriate corresponding threshold values and foresee the embedding results under this threshold value. We extend our prediction algorithm to an image database consisting of a number of cover images. This allows us to recommend the best cover image from the image database which has the highest embedding performance or satisfies user's need for concealing secret messages. Experimental results show that our mechanism has high prediction accuracy. The average error rate between the capacity being predicted and the capacity derived from practical embedding is less than 10^(-6)%. The average error rate between the peak signal-to-noise ratio being predicted and the real one is less than 10^(-6)%. In conclusion, our prediction mechanism provides five significant features; namely, prediction, convenience, optimization, generality, and integration. The scheme is feasible to steganographic applications.
URI: http://hdl.handle.net/11455/72818
ISSN: 1017-4397
Appears in Collections:第23卷 第3期
工學院

Files in This Item:
File SizeFormat Existing users please Login
10174397-201211-201212060037-201212060037-119-144.pdf4.33 MBAdobe PDFThis file is only available in the university internal network   
Show full item record
 

Google ScholarTM

Check


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