透過您的圖書館登入
IP:3.138.33.87
  • 學位論文

使用多位元浮水印回復幾何變形影像

Image Recovery of Geometric Distortion Using Multi-bit Watermarking

指導教授 : 李明穗

摘要


本論文提出了一種新的浮水印技術能使用在回復幾何變形影像與攜帶資訊。 在我們的浮水印技術中,許多樣板會被藏在小波轉換域中。這些樣板能提供回復 幾何變形影像時所需的參考資訊。我們提出了一個演算法能在幾何變形的影像中 找出樣板的位置與所藏的資訊,值得注意的是我們的演算法並不需要原始圖片的 支援。此演算法是建構在貝氏網路的架構上。在幾何變形影像中樣板的位置和所 藏的資訊能夠以機率的方式表示在貝氏網路中。藉由推論貝式網路,樣板的位置 和所藏的資訊可以被估測出來。估測出樣板的位置後,樣板的位移就得以被計算 出來。接著再使用差補法估測出每個像素的位移,如此一來便能藉由這些資訊將 幾何變形影像回復。實驗的結果顯示出我們的浮水印技術能很強健的對抗幾何變 形,並且將幾何變形影像復原。

並列摘要


A novel watermarking scheme which recovers an image from geometric distortion and carries multi-bit binary message is proposed in this thesis. In our watermarking scheme, several templates are inserted into the discrete wavelet transform domain and serves as the registration references. We provide a blind watermarking extraction algorithm to extract the hidden message and locate the templates. The locations of templates and binary message are represented probabilistically in the Bayesian network. The displacement parameters of each template can further be obtained once the locations of the templates are estimated. The distorted image is recovered based on the displacements of each pixel which is estimated by interpolation. Experiments demonstrate that proposed watermarking scheme are robust to local geometric distortion and is capable of recovering distorted images well.

參考文獻


search and the false positive watermark detection probability,” SPIE, Security and
watermarking robust to geometric distortions,” IEEE Transactions on Image
Processing, pp. 2140–2150, 2005.
[4] D. Coltuc and P. Bolon, “Robust watermarking by histogram specification,” IEEE
International Conference of Image Processing, pp. 236–239, October 1999.

延伸閱讀