Aiming at the security protection of remote sensing images, a robust watermarking algorithm based on SURF (Speeded Up Robust Features) feature on selective regions is proposed. The algorithm first extracts the SURF feature points of the carrier, and then performs a 5/3 integer wavelet transform on the carrier image to filter out the low-frequency coefficients of the ROI and the intermediate frequency coefficients of the non-interest area (ROB); With sampling pyramid decomposition, the near subband after watermark decomposition is embedded in the lowfrequency subband of the region of interest, and the residual subband is embedded in the intermediate frequency coefficient of the non-interesting region. Experimental data show that the algorithm can resist conventional geometric attacks. The similarity of the watermark is high, and the NC value is kept above 0.89, which has good reversibility and robustness.
Remote sensing imagery is an important carrier of geospatial information, and its military and economic value is increasingly prominent, and it plays an important role in many fields such as surveying and mapping, navigation, reconnaissance, and monitoring[1]. However, the digital storage method and open network environment not only realize the rapid transmission and efficient sharing of remote sensing images, but also bring new challenges to the security protection of image data. In recent years, data leakage, illegal tampering, and ownership violations against remote sensing images have been repeatedly prohibited. Digital watermarking technology is a cuttingedge technology developed in the field of information security and an important means of remote sensing image security protection.
Although the research of remote sensing image digital watermarking technology started relatively late, it has also achieved vigorous development due to its great practical significance. Bhatti et al. [2] studied the evaluation criteria of digital watermarking for high resolution color images, and pointed out that watermarking technology for ordinary images is not completely suitable for medical images images and can be used by separating the color images in RGB color space. Delaigle et al. [3] uses human visual characteristics and visual models to select important wavelet coefficients to embed the watermark, but the original image is required to participate in the detection, which is a non-blind algorithm and is not practical.
Saqib et al. [4] embeds the encrypted binary image watermark into the block-scrambling remote sensing image, which has good robustness to conventional attacks, but cannot resist geometric attacks. Pereira et al. [5] uses the template matching method to resist geometric attacks, but the key matrix is required to participate in the detection, which is a semi-blind algorithm. The literature [6] embeds the watermark into the normalized remote sensing image in the Controllet domain, but because it is embedded as a whole, the algorithm is not robust to the cutting of the image size. In general, the current research on remote sensing image watermarking algorithms mostly uses the first-generation watermarking method [2-7], and rarely involves the secondgeneration watermarking technology, that is, algorithms based on image features. However, in specific applications, remote sensing images embedded with watermarks inevitably need to be rotated, zoomed, cropped to change the original size, and tile stitching, etc., and the angle of rotation, zoom multiples, etc. during watermark detection The position relative to the original image after cutting and splicing is unknown. These geometric attacks destroy the synchronization of the watermark, resulting in detection failure. Algorithms based on image features provide a brand-new idea for solving this problem, and the research on algorithms for ordinary images has been relatively in-depth [8-13], which can provide methodological references for the research of remote sensing image watermarking.
Based on the existing algorithms, this paper designs a robust blind watermarking algorithm for remote sensing images based on SURF feature points and the excellent characteristics of region based feature selection using region of interest (ROI), which is strong against conventional attacks and geometric attacks. The main contributions of this study are: 1) Secure watermarking algorithm for security of remote sensing images.
2) Implementation of SURF using feature based region selection for watermarking.
SURF (speeded up robust features) is a fast-robust local feature detection algorithm proposed Based on SIFT operator. In general, the standard SURF operator is several times faster than the SIFT operator and has better robustness under multiple images [14]. This paper uses ROI selection based on SURF features. The basic idea is as follows: first, calculate the integral image and traverse the image once to get the sum of all pixels. Then construct the Hessian matrix [15] and perform Gaussian filtering on the image. After filtering, the Hessian matrix expression is: =
L L ((x, y) , σ) ((x, y) , σ)
L
L
((x, y) , σ)
((x, y) , σ)
(
If the Hessian matrix discriminator has an extreme value, the current point will be brighter or darker than the surrounding points, and the candidate object may be divided by the extreme value. To increase the speed, SURF uses a box filter to approximate a Gaussian filter. If the endpoint is = − µ ( ) µ ( )
Where represents the d (H) value of the feature point p, and µ ( ) is the average of the d (H) values of 26 points around the point p. The matrix composed of the contribution of feature points is the contribution matrix. Using the idea of dynamic programming to determine the largest sub-matrix, the matrix is the part with the largest contribution of feature points, that is, ROI. SURF feature point correction: let (X , Y ) and (X , Y ) be any two feature points in the original image feature points, (X′ , Y′ ) and (X′ , Y′ ) are Feature points of image matching after suffering a geometric attack.
Rotation correction: If the number of matching feature points is N, then the angle between the vectors of the matching feature points of the two images is the angle of rotation. From the vector angle formula (3), the maximum rotation angle is removed. The obtained angle is averaged to obtain the rotation angle β. a physical endpoint, it is very important to calculate the Hessian discriminant for each pixel. If it is a positive number, the pixel is a local extreme point, otherwise, it is not. The extreme point is obtained is used as a candidate feature point. Then, the non-maximum suppression of the 3 * 3 * 3 cube neighborhood adjacent to this point [16], that is, the candidate extremum point is related to 8 extremum points of the same scale neighborhood and 18 extremum points of the adjacent scale. In comparison, the higher the significance of the pixel and the greater the contribution to the ROI selection. The feature point contribution is defined as (2) (3) (4) (5) (6)
Scaling correction: According to the matching feature points of the two images, the scaling ratio of the image length and width can be estimated, and the points with larger errors can be removed, and the scaling ratio of length and width can be obtained by averaging.
Translation correction: Calculate the difference between the abscissa and ordinate of each pair of matching feature points of the two images, remove the larger error value, and calculate the average value to get the translation distance.
= ( )(
) 1 =
≤ − 1, ≤ − 1 = ,
= ∆ = | − | ∆ = |
− |
The digital watermark is sampled, and the residual sub-band is calculated to generate a sampling golden tower composed of a series of residual sub-bands and an approximate sub-band. The image of this golden tower structure has scalable characteristics. Set the original image as the bottom layer of the sampling pyramid (layer 0), down-sampling to obtain the first layer of the sampling golden tower, and then fill the with the interpolation method to form the same as the original image The size of the image ∗. Then the difference between and ∗ is used to construct the residual image . After decomposing the sampled golden tower in one layer, an approximate image and a residual image are generated. If the sampling gold tower decomposition needs to be continued, a similar operation is performed on the approximate subband image to generate an approximate image
and a residual image . ∗( , ) = 4 ( , )
(7) + + 2 2
Among them: of ROI and ROB.
The image is composed of 5/3 3-level IWT to decompose the extract of wavelet coefficients (3) Arnold scrambling of watermark and three-level sampling pyramid decomposition to obtain 4 subband data: , , and , where is approximate subband, , and are the third level, second Level, and the first level residual subband.
(4) The approximate
sub-band watermark is decomposed into a sampling pyramid, and then ROI is embedded in the LL sub-band using the reversible watermark histogram algorithm. (5) The residual subbands , and
of the watermark are embedded into LH , LH and
LH of ROB through singular value decomposition. Embedding method: After each h-h block is
divided into h × h blocks, SVD decomposition is performed, A = USV , and Q = round (S (
Embed the watermark according to equation (10).
∗ , = 0 , is an integer (8)
When reconstructing the image sampling pyramid, from the top to the bottom of the sampling golden tower, the following formula is used to restore layer by layer, and then the original image is obtained.
When reconstructing the image sampling pyramid from the top to the bottom of the tower, use the following formula to copy the sampled gold layer by layer, and then save the original image. = = + ∗ 0 ≤ ≤ =
The specific steps of watermark embedding are shown in Figure 1. Extract the SURF feature points of the carrier image, as described in Section 1.1, select the image ROI according to the part with a large contribution of feature points; the wavelet subband data of the ROI.
(9) (10) subband information , , embedding method, and SVD.
(4) Using the singular value decomposition algorithm to extract the watermark residual
of LH , LH , LH in ROB. The extraction method is similar to the
decomposition is performed on each h × h block, A = US
and calculate d =
floor (S (
= 1 0 (5) Perform sampling pyramid reconstruction on the watermark subband information extracted in step (3) and step (4). Then, the inverse Arnold transformation is performed on the reconstructed image to obtain the extracted watermark.
The experimental environment is MATLAB2018, which performs invisibility test, multiresolution extraction test and robustness test respectively. The experimental carrier is 512 × 512 remote sensing image, and the watermark is 32 × 32 binary image. Fig. 5 shows that remote sensing image and the method of watermark embedding.
The carrier images of the experiment are remote sensing image. The embedding intensity of the watermark in ROB and remote sensing image is obtained after embedding the watermark, as shown in Figure 3, and the peak signal-to-noise ratio (PSNR) is shown in Table 1. The Gaussian noise 4% and JPEG 20% shows that NC is 0.93 and 1 after extraction, and the robustness is very good.
The NC PSNR under Conventional Attacks .
0.92
4% 14.65 0.93
6% 13.18 0.93
10% 20% 25.57 28.66 0.91 1
50% 32.54 1
According to the algorithm proposed in this paper, the image under attack is corrected and then the watermark is extracted, and the image is only rotated, respectively. The rotation angle of the test is set to 10 . −50 . The algorithm first calculates the difference of the original image, calculates the difference histogram of the image and finds the peak value, and embeds the watermark through the peak value.
. Table 2 and Figure 6 shows that the results against different attacks: For geometric attacks such as translation, rotation, and scaling, the NC values extracted by the algorithm in this paper are all above 0.81, and the NC value can be 1 when the rotation angle anticlockwise is 10° and translation down 10%. It can be seen that all attack results of NC value are good.
Table2.PSNR and NC under Geometric Attacks.
Geometric Attacks Attack strength PSNR（dB） NC 10⁰ 11.62 0.86 (Rcol otactkiownise) 30⁰ 10.83 1 50⁰ 10.33 0.77 10⁰ 11.51 1 Rotation 30⁰ 10.69 0.81 (Anticlockwise) 50⁰ 10.41 0.88 Scaling ˣˣ 00..68 -- 00..7851
10% 10.74 0.84 Tr(aRnisglhatti)on 3200%% 91.07.822 00..8844 Tra(dnoslwatni)on 1300%% 19.18.227 10.80
50% 8.67 0.74 Clipping 10% - 1 (Y direction) 30% - 0.84 (CXlipdpiriencgtion) 1300%% -- 00..9933
Fig 4: Different attacks on remote sensing image
This paper adopts the idea of the second generation of watermarking, combining the excellent characteristics of SURF operator and integer wavelet transform, taking into account the characteristics of remote sensing images, and proposes a robust blind watermarking algorithm for remote sensing images based on SURF feature regions. While maintaining the accuracy of remote sensing image data, the algorithm can effectively resist conventional attacks such as noise, filtering, JPEG compression, brightness adjustment, and geometric attacks such as rotation, scaling, cutting, and stitching, without the need to correct and restore the attacked image. Watermark can be extracted from it, which has strong practicability and efficiency, and can effectively protect the security of remote sensing images. [2] Bhatti, U. A., Yu, Z., Li, J., Nawaz, S. A., Mehmood, A., Zhang, K., & Yuan, L. (2020).
Hybrid watermarking algorithm using clifford algebra with Arnold scrambling and chaotic encryption. IEEE Access, 8, 76386-76398. [3] Delaigle, J. F., Devleeschouwer, C., Macq, B., & Langendijk, L. (2002, August). Human visual system features enabling watermarking. In Proceedings. IEEE International Conference on Multimedia and Expo (Vol. 2, pp. 489-492). IEEE.. [4] Nawaz, S. A., Li, J., Bhatti, U. A., Mehmood, A., Shoukat, M. U., & Bhatti, M. A. (2020).
Advance hybrid medical watermarking algorithm using speeded up robust features and discrete cosine transform. Plos one, 15(6), e0232902 [5] Pereira, S., & Pun, T. (2000). Robust template matching for affine resistant image watermarks. IEEE transactions on image Processing, 9(6), 1123-1129. [6] Jing, L., Zhang, Y., & Chen, G. (2008, October). Zero-watermarking for copyright protection of remote sensing image. In 2008 9th International Conference on Signal Processing (pp. 1083-1086). IEEE. [7] Cayre, F., Fontaine, C., & Furon, T. (2005). Watermarking security: theory and practice.
IEEE Transactions on signal processing, 53(10), 3976-3987. [8] Jiansheng, M., Sukang, L., & Xiaomei, T. (2009). A digital watermarking algorithm based on DCT and DWT. In Proceedings. The 2009 International Symposium on Web Information Systems and Applications (WISA 2009) (p. 104). Academy publisher. [9] Kang, X., Huang, J., Shi, Y. Q., & Lin, Y. (2003). A DWT-DFT composite watermarking scheme robust to both affine transform and JPEG compression. IEEE transactions on circuits and systems for video technology, 13(8), 776-786. [10] Bhatnagar, G., & Raman, B. (2009). A new robust reference watermarking scheme based on DWT-SVD. Computer Standards & Interfaces, 31(5), 1002-1013. [11] Al-Haj, A., Mohammad, A. A., & Bata, L. (2011). DWT-based audio watermarking.
Int. Arab J. Inf. Technol., 8(3), 326-333. [12] Thakkar, F. N., & Srivastava, V. K. (2017). A blind medical image watermarking: DWT-SVD based robust and secure approach for telemedicine applications. Multimedia Tools and Applications, 76(3), 3669-3697. [13] Gupta, P., & Parmar, G. (2017, July). Image Watermarking using IWT-SVD and its Comparative Analysis with DWT-SVD. In 2017 International Conference on Computer, Communications and Electronics (Comptelix) (pp. 527-531). IEEE. [14] Alotaibi, R. A., & Elrefaei, L. A. (2019). Text-image watermarking based on integer wavelet transform (IWT) and discrete cosine transform (DCT). Applied Computing and Informatics, 15(2), 191-202. [15] Solanki, N., Malik, S. K., & Chhikara, S. (2014). Roni medical image watermarking using dwt and rsa. International Journal of Computer Applications, 96(9). [16] Ayubi, P., Barani, M. J., Valandar, M. Y., Irani, B. Y., & Sadigh, R. S. M. (2021).
A new chaotic complex map for robust video watermarking. Artificial Intelligence Review, 54(2), 1237-1280. [17] Nazari, M., & Mehrabian, M. (2021). A novel chaotic IWT-LSB blind watermarking approach with flexible capacity for secure transmission of authenticated medical images. Multimedia Tools and Applications, 80(7), 10615-10655.