Method for embedding and extracting digital watermark on lowest wavelet subband

• US 7,343,025 B2
• Filed: 12/15/2003
• Issued: 03/11/2008
• Est. Priority Date: 10/02/2003
• Status: Expired due to Fees

• Current Assignee: Electronics and Telecommunications Research Institute
• Original Assignee: Electronics and Telecommunications Research Institute
• Inventors: Joo, Sanghyun, Lee, Seon Hwa, Seo, Yong Seok
• Primary Examiner(s): ALLISON, ANDRAE S
• Application Number:
  US10/734,142
• Publication Number:
  US 20050074139A1
• Time in Patent Office: 1,548 Days
• Field of Search: 382/100, 382/240, 382/253, 382/232, 382/250, 386/94, 380/253
• US Class Current: 382/100
• CPC Class Codes: G06T 1/0057 Compression invariant watermarking
• CPC Class Codes: H04N 1/3217 using wavelet transforms
• CPC Class Codes: H04N 1/32187 with selective or adaptive application of the additional information, e.g. in selected frequency coefficients

Abstract

A for embedding a digital watermark on a wavelet lowest subband includes: setting a DC component region of a multi-stage wavelet-transformed original copy image to a watermark embedment region, and high-frequency filtering an original picture LL_n of the embedment region; generating index information for designating a pixel position, and a watermark sequence to be embedded; calculating an embedment strength λ for each position of the watermark embedment region; in case the watermark sequence is sequentially embedded on an embedded position designated by the index information, mutually comparing the original picture LL_n coefficient value for each embedded position with a mirror picture LL_n′ coefficient value, and then altering the original picture LL_n coefficient value; and in case the original picture LL_n coefficient value is differentiated above a predetermined value with reference to the coffesponding embedment strength λ, skipping the watermark embedment for the position.

Claims

1. A method of embedding a digital watermark on a wavelet lowest subband of an original image, the method comprising the steps of:

wavelet transforming the original image into n levels;

setting the wavelet lowest subband of the wavelet-transformed original image as a watermark embedment region;

high-frequency filtering an original picture LL_n of the watermark embedment region to generate a mirror picture LL_n′

from which a high frequency component of the original picture LL_n has been eliminated;

generating (i) index information designating a plurality of pixel positions within the watermark embedment region where the watermark is to be embedded, and (ii) a sequence of watermark values to be embedded in said pixel positions, respectively;

calculating an embedment strength λ

for each said pixel position of the watermark embedment region, considering a variance degree of an original picture LL_n coefficient value;

for each said pixel position, selectively altering the original picture LL_n coefficient value based on (a) a corresponding mirror picture LL_n′

coefficient value, (b) the watermark value to be embedded in said pixel position, and (c) the embedment strength λ

of said pixel position; and

selectively embedding the watermark at each said pixel position with the selectively altered original picture LL_n coefficient value while skipping watermark embedment where the original picture LL_n coefficient value before and after said altering differs by more than a predetermined value associated with the corresponding embedment strength λ

.

2. The method of claim 1, wherein in the step of high frequency filtering, the high-frequency component is eliminated though Wiener filtering.

3. The method of claim 1, wherein in the embedment strength calculating step, the embedment strength λ

in every said pixel position in the watermark embedment region is calculated according to the following Equations (1) and (2);

$\begin{array}{cc}\mathrm{NVF}<br><br>\left(i,j\right)=\frac{{\mathrm{\sigma <br><br>}}_{\max }^2}{{\mathrm{\sigma <br><br>}}_{\max }^2+\mathrm{\theta <br><br>}<br><br><br><br>{\mathrm{\sigma <br><br>}}^2<br><br>\left(i,j\right)}& \left(1\right)\\ \mathrm{\lambda <br><br>}<br><br>\left(i,j\right)=S_e·<br><br>\left(1-\mathrm{NVF}<br><br>\left(i,j\right)\right)+S_f·<br><br>\mathrm{NVF}<br><br>\left(i,j\right)& \left(2\right)\end{array}$

σ

²(i, j);

a local variance value for a peripheral region centering on the position (i,j);

σ

²_max;

a maximum local variance value in the watermark embedment region;

S_e, S_f and θ

;

embedment strength controlling values.

4. The method of claim 1, wherein, in the step of selectively altering the original picture LL_n coefficient value, if the watermark value ‘

1’

is to be embedded in a corresponding pixel position i, the original picture coefficient value LL_n(i) is compared with an added value LL_n′

(i)+λ

(i) obtained by adding the embedment strength to the mirror picture coefficient value and

if the original picture coefficient value is greater than the added value, the original picture coefficient value is maintained as is, and

if the original picture coefficient value is less than the added value, the original picture coefficient value of the pixel position is assigned the added value.

5. The method of claim 1, wherein, in the step of selectively altering the original picture LL_n coefficient value if the watermark value ‘

−

1’

is to be embedded in a corresponding pixel position i, the original picture coefficient value LL_n(i) is compared with a subtracted value LL_n′

(i)−

λ

(i) obtained by subtracting the embedment strength from the mirror picture coefficient value and

if the original picture coefficient value is less than the subtracted value, the original picture coefficient value is maintained as is, and

if the original picture coefficient value is greater than the subtracted value, the original picture coefficient value is assigned the subtracted value.

6. The method of claim 1, wherein the initial original picture efficient value is maintained to skip the watermark embedment at each pixel position where the original picture LLn coefficient value before and after said altering differs by more than thee times the embedment strength.

7. The method of claim 1, wherein the watermark is repetitively embedded in each pixel position a predetermined number of times depending on robustness and screen degradation degree.

8. A watermark extracting method of extracting a watermark embedded by the method of claim 1 from a watermark embedded image, the watermark extracting method comprising the steps of:

wavelet transforming the watermark embedded image into the same n levels;

defining a wavelet lowest subband of the wavelet-transformed watermark embedded image as a watermark extracted region;

high-frequency filtering an original picture LL_nE of the watermark extracted region to generate a mirror picture LL_n′

from which a high frequency component of the original picture LL_nE has been eliminated;

receiving index information designating a plurality of pixel positions within the watermark extracted region where the watermark has been embedded;

comparing an original picture LL_nE coefficient value with a miffor picture LL_nE′

coefficient value at each said pixel position to extract a sequence W_E(i) of watermark values;

receiving a key value to regenerate an original sequence W(i) of watermark values used at the time of watermark embedment; and

determining a similarity between the extracted watermark sequence and the original watermark sequence, and judging whether or not the watermark exists depending on whether or not the similarity is greater than a predetermined critical value.

9. The method of claim 8, wherein, in the watermark extracting step,

if the original picture LL_nE coefficient value is less than the mirror picture LL_nE′

coefficient value, a watermark value ‘

−

1’

is extracted from the respective pixel position;

otherwise, a watermark value ‘

+1’

is extracted.