1 / 22

Robust Invisible Watermarking of Volume Data Using the 3D DCT

Robust Invisible Watermarking of Volume Data Using the 3D DCT. Yinghui Wu , Xin Guan,Mohan S.Kankanhalli,and Zhiyong Huang. Outline. The introduction of spread-spectrum The volume watermarking technique Watermark detection Test results. The introduction of spread-spectrum.

abel
Download Presentation

Robust Invisible Watermarking of Volume Data Using the 3D DCT

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Robust Invisible Watermarking of Volume Data Using the 3D DCT Yinghui Wu , Xin Guan,Mohan S.Kankanhalli,and Zhiyong Huang

  2. Outline • The introduction of spread-spectrum • The volume watermarking technique • Watermark detection • Test results

  3. The introduction of spread-spectrum • Least significant bit • Direct sequence spread • Block spread spectrum • Duplication spreading

  4. Least significant bit • Least significant bit embedding • 第一步：將影像資料轉換為8bit的二進位明文，舉例如果灰色圖素的值是90轉換結果值將為01011010。 • 第二步：選擇圖素的Least significant bit 插入watermark一個位元轉換成新的資料，舉例說明如果我們想要嵌入一個”1”在灰色值90的圖素裡我們將得到一個新的圖素資料91它的二進位是01011011。 當我們執行 L S B數位浮印技術我們可以得到的好處是非常的簡單快速而且容易製作，加入浮水印的位元設在圖片區塊位元的最低位元,是不易被人眼所觀察出來的。但是相對的它的缺點是容易被雜訊及幾何改變的破壞，容易被刪除，安全性不高。

  5. Direct sequence spread • 步驟一：假設watermark的資料流是”1011001”，原始圖檔資料流是”00101001(41)，01010100(84)，00111010(58)，10000111(135)，00011111(31)，10001000(136)，00000000(0)，11111111(255)，10101111(175)，…..” • 步驟二：將watermark的資料流”1011001”延展3次可得”111000111111000000111”。 • 步驟三：如果選擇 least significant bit 的方法去嵌入將可得到”00101010(42)，01010101(85)，00111011(59)，10000111(135)，00011111(31)，10001001(136)，00000001(1)，11111111(255)，10110000(176)，…”。 • 步驟四：將嵌入的資料取出我們可得到”1，1，1，0，0，0，1，0，1…”。 • 步驟五：應用多數理論的原則來還原原始的watermark得到的是”1，0，1，1，0，0，1”。

  6. Block spread spectrum • 如果利用least significant bit 將logo X嵌入在圖檔Y裡面，這個資料的容量等於圖檔圖素的總合，典型的區段展頻 least significant bit 資料嵌入的作法如下步驟 • 步驟一：假設logo X是I*J，圖像資料Y的容量是M*N，I，J << M，N(最好>32次)。步驟二：展開X的容量I*J變成二位元的資料流，得到I*J*8 位元資料 X’。 • 步驟三：將圖像資料Y分割成(M*N)/(I*J*8)個區塊，我們叫做Y’區塊陣列。 • 步驟四：每一個Y’的區塊陣列圖素的least significant bit 位置上嵌入X’資料，一次一個bit，順序是Y’[1，1]，Y’[1，2]…Y’[k，n-k]。 • 以上的方法是利用區段展頻結合least significant bit，圖檔的容量剛好等於圖素的總合，在這裡，圖檔將被區分為多少個區塊呢?這個答案將根據不同的logo和不同的圖檔特性來決定。

  7. Duplication spreading • 重複展頻的基礎只是應用反覆的嵌入處理，將logo展開的二位元資料流嵌入圖檔的每一個圖素裡，重複執行，直到圖檔的最後一個圖素。作法如下 • 步驟一：假設logo X的容量是I*J，圖檔Y的資料容量是M*N，在這裡 M，N >= I，J。 • 步驟二：展開logo X成為二位元的資料，得到結果是 I*J*8位元資料流 X’。步驟三：將X’嵌入圖檔Y順序是Y[1，1]，Y[1，2]…Y[M，N]，按照順序重複X’直到Y檔案結束為止。 • 應用重複展頻技術可提高資料還原及安全性，此方法僅較優於直接序列展頻。

  8. The volume data watermarking technique • We utilize the spread-spectrum technique in the frequency domain in order to achieve this effect. • Assume that volume v that needs to be watermarked is of the size .

  9. The volume watermarking technique (cont.) • A block-based 3D DCT transform is applied to the volume V. where f(x,y,z) corresponds to the voxel values,and F(u,v,w) corresponds to the 3D DCT coefficient.

  10. The volume watermarking technique (cont.) • To embed the watermark information bits the bits are first spread by a large spread factor cr, called the chiprate. The spreading provides spatial redundancy by enbdeeing the information bits into cr number of voxels and K varies form 1 to cr.

  11. The volume watermarking technique (cont.) • The spread bits are then modulated with a pseudo-random-noise(PN) sequence. This form the basic watermark sequence • The watermark sequence , forms a volume W of size .

  12. The volume watermarking technique (cont.) • For every DCT block and the corresponding DCT block , the corresponding coefficients are added to form a watermarked block which constitute the watermarked volume in the frequency domain.

  13. The volume watermarking technique (cont.) • The 3D inverse DCT is performed on to obtain a size volume V’.

  14. Watermark detection • For detecting the existence of the watermark , the DCT-transformed original volume data watermarked volume data is subtracted from the DCT-transformed watermarked volume data to obtain the residual volume data DCT coefficients , i.e. • The 3D inverse DCT is performed on this residual data to obtain the residual watermark sequence

  15. Watermark detection (cont.) • Considering one subset of the watermark values over the correlation window where where being the error term which can be due go intentional or unintentional attacks. • By choosing a large cr we have adequate redundancy and the summation can be approximated as : The required information bit :

  16. Test results • In order to verify the robustness and invisible property of the algorithm proposed , we used a skull dataset (68*64*64) and a larger tomato data set(64*208*216) to conduct tests.

  17. Test results (cont.)

  18. Test results (cont.) • The distortion caused by the attacks is measured in terms of SNR (Signal-to-Noise) and PSNR (Peak-Signal-to-Noise Ratio): where and the ith-voxel values of the original and watermarked volume data respective.

  19. Test results (cont.) • First we conducted the cropping attacks. Experiment results of robustness on volume cropping is shown in table 1 for the skull dataset. watermark length=1840 bits and the chiprate cr=53

  20. Test results (cont.) • We did the test of table 2 using a 2D image of “NUS” logo as the watermark.

  21. Test results (cont.) • The Fig.3 shows the retrieved watermarks (2D images) under the various noise levels.

  22. Test results (cont.) • For table 3 of the tomato dataset. Watermark length l=4000bits and the Chiprate cr=719.

More Related