Digital watermarking has gained importance in recent old ages in right of first publication protection and multimedia hallmarks. In this paper an effort is made to analyze the attack of combined DWT-DCT watermarking technique on orbiter and bio-medical images. The method provides improved imperceptibility and hardiness to the screen images. In the first measure, the screen image is decomposed into two degrees by DWT transform. Then DCT of the HL ( or HH ) sub set of the DWT coefficients is computed. The water line is embedded in the signifier of a PN ( pseudo random ) sequence into the DWT bomber set after calculating DCT. The technique is tested on screen image LENA ( 512×512, 8-bit grey graduated table ) with a binary water line of size 50×20. The watermarked images are tested for hardiness against JPEG compaction and other image processing onslaughts. Experimental consequences show that uniting the two transforms improved the public presentation when compared to merely DWT transform methods.
Digital watermarking refers to techniques used to protect digital informations by unnoticeably implanting information ( water line ) into the original informations in such a manner that it ever remains present. For a digital watermarking method to be effectual, it should be unperceivable and robust to common image uses such as compaction, filtering, rotary motion, grading, and cropping onslaughts among many other digital signal processing operations. Current digital image watermarking techniques can be grouped in to two major categories: spatial-domain and frequence sphere techniques. Frequency domain watermarking techniques have proved to be more effectual with regard to accomplishing the imperceptibility and hardiness demands of digital watermarking algorithms.
Normally used frequency-domain transforms are the Discrete Wavelet Transform ( DWT ) , the Discrete Cosine Transform ( DCT ) and Discrete Fourier Transform ( DFT ) . However, DWT has been used in digital image watermarking more often due to its first-class spacial localisation and multi-resolution features, which are similar to the theoretical theoretical accounts of the human ocular system ( HVS ) . Further public presentation betterments in DWT-based digital image watermarking algorithms could be obtained by uniting DWT with DCT. This is based on the fact that combined transforms could counterbalance for the drawbacks of each other, ensuing in effectual watermarking. Watermarking is done by changing the ripples coefficients of selected DWT sub sets ( HL or HH ) , followed by the application of DCT transform on selected bomber sets.
2. DCT and DWT TRANSFORMS
The DCT and DWT transforms have been extensively used in many digital signal processing applications. In this subdivision, we introduce the two transforms and sketch their relevancy to the execution of digital watermarking.
2.1. DCT Transform: The distinct cosine transform ( DCT ) is a technique for change overing a signal into simple frequence constituents. It represents an image as a amount of sinusoids of changing magnitudes and frequences. With an input image degree Fahrenheit, the DCT coefficients for the transformed end product image, T, are computed harmonizing to Eqn.1 given below. Here, degree Fahrenheit, is the input image with MxN pels, degree Fahrenheit ( m, N ) is the strength of the pel in row m and column N of the image, T ( u, V ) is the DCT coefficient in row U and column V of the DCT matrix.
( 1 )
The image is reconstructed by using reverse DCT operation harmonizing to equation 2 ) .
( 2 )
degree Fahrenheit ( x, y ) is the original screen image and
F ( x, Y ) is the watermarked image
The block-based DCT transform sections an image in to non overlapping blocks and applies DCT to each block. This will give three frequence sub sets: low frequence bomber set, mid-frequency bomber set and high frequence bomber set. DCT watermarking is based on two facts. The first is that much of the signal energy prevarications at low frequences sub set which contains the most of import ocular parts of the image. The 2nd fact is that the high frequence constituents of the image are normally removed through compaction and noise onslaughts. The water line is hence embedded by modifying the coefficients of the in-between frequence bomber set so that the visibleness of the image will non be affected and the water line will non be destroyed by compaction.
Fig.1. Watermark implanting process utilizing combined DWT-DCT method
2.2 DWT Transform:
For 2-D images, using DWT corresponds to treating the image by 2-D filters in each dimension. The filters divide the input image into four non-overlapping multi-resolution bomber sets: LL1, LH1, HL1 and HH1. The sub set LL1 represents the coarse-scale DWT coefficients while the sub sets LH1, HL1 and HH1 represent the fine-scale DWT coefficients. To obtain the following coarse graduated table ripple coefficients, the bomber set LL1 is farther decomposed.
Due to its first-class spacial frequence localisation belongingss, the DWT is really much suitable to place the countries in the screen image where a water line can be embedded efficaciously. In peculiar, this belongings allows the development of the dissembling consequence of the HVS such that if a DWT coefficient is modified, merely the part matching to that coefficient will be modified. In general most of the image energy is concentrated at the lower frequence bomber bands LL, and hence implanting water lines in these sub sets may degrade the image significantly. Implanting in the low frequence bomber sets, nevertheless, could increase hardiness significantly. On the other manus, the high frequence bomber bands HH include the borders and textures of the image and the human oculus is non by and large sensitive to alterations in such sub sets. This allows the water line to be embedded without being perceived by the human oculus. The via media adopted by many DWT-based watermarking algorithm, is to implant the water line in the in-between frequence bomber bands LH and HL where acceptable degree of imperceptibility and hardiness could be achieved.
( a )
( B )
Fig.2. Two Level DWT decomposition of the original image
3. DWT-DCT METHOD
3.1. Watermark Implanting
The water line embedding is performed with the undermentioned stairss ( Fig.1 )
Measure 1: Read in the screen image degree Fahrenheit ( x, y ) .
Measure 2: Apply DWT to break up the degree Fahrenheit into four non-overlapping multi-resolution bomber sets: LL1, HL1, LH1, and HH1.
Measure 3: Apply DWT to stand in set HL1 to acquire four smaller bomber sets: LL2, HL2, LH2, and HH2. Choose the HL2 bomber set ( Fig. 2a ) , or use DWT to stand in set HH1 to acquire four smaller bomber sets: LL2, HL2, LH2, HH2. Choose HH2 bomber set as shown in Fig. 2b.
Measure 4: Divide the bomber set HL2 ( or HH2 ) into 4×4 blocks.
Measure 5: Perform DCT on each 4×4 block in the chosen bomber set ( HL2 or HH2 ) .
Measure 6: Read in the binary water line image.
Measure 7: Generate two uncorrelated pseudorandom ( PN ) sequences. One sequence is used to implant the water line spot 0 ( PN0 ) and the other sequence is used to implant the water line spot 1 ( PN1 ) . The figure of elements in each of the two pseudorandom sequences must be equal to the figure of mid-band elements of the DCT-transformed DWT bomber sets.
Measure 8: Embed the pseudorandom sequences PN0, PN1, with addition factor K in the DCT transformed 4×4 blocks of the selected DWT sub sets of the screen image. Embedding is applied merely to the mid-band DCT coefficients. If X is the matrix of the mid set coefficients of the DCT transformed block, so implanting is done as follows:
If the water line spot is 0 so,
X? = X + k * PN_0 ( 3 )
Otherwise, if the water line spot is 1 so,
X? = X + k * PN_1 ( 4 )
Where X? is the watermarked DCT block
Measure 9: Apply inverse DCT ( IDCT ) to each 4×4 block after its mid-band coefficients have been modified by implanting the water line spots.
Measure 10: Use the opposite DWT ( IDWT ) to bring forth the watermarked screen image.
3.2 Watermark Extraction
The water line extraction process is shown in Fig. 3, and described in item in the undermentioned stairss. In the proposed combined DWT-DCT algorithm watermarking algorithm, the original screen image is non required to pull out the water line.
Measure 1: Read in the watermarked screen image
Measure 2: Apply DWT to break up the watermarked image into four non-overlapping multi-resolution bomber sets: LL1, HL1, LH1, and HH1
Measure 3: Apply DWT to HL1 to acquire four smaller bomber sets, and take the bomber set HL2, as shown in Fig. 2 a. or, apply DWT to the HH1 bomber set to acquire four smaller bomber sets, and take the HH2 bomber set, as shown in Fig. 2b.
Measure 4: Divide the bomber set HL2 ( or HH2 ) into 4?4 blocks.
Measure 5: Apply DCT to each block in the chosen bomber set ( HL2 or HH2 ) , and pull out the mid-band coefficients of each DCT transformed block.
Measure 6: Regenerate the two pseudorandom sequences ( PN0 and PN1 ) utilizing the same seed used in the water line implanting process.
Measure 7: For each block in the bomber set HL2 ( or HH2 ) , calculate the correlativity between the mid-band coefficients and the two generated pseudorandom sequences ( PN0 and PN1 ) . If the correlativity with the PN0 was higher than the correlativity with PN1, so the extracted water line spot is considered 0, otherwise the extracted water line is considered 1.
Measure 8: Reconstruct the water line utilizing the extracted water line spots, and calculate the similarity between the original and extracted water lines.
Fig.3. Watermark extraction process
4. PERFORMANCE EVALUATION
The public presentation of combined DWT-DCT image watermarking algorithms is evaluated on the screen image ‘HEAD ‘ with a 50×20 binary image ‘SVEC ‘ as the water line image. The two images are shown in Fig. 4 and 5, severally. Watermarking algorithms are normally evaluated with regard to two prosodies: imperceptibility and hardiness [ 13 ] .
Imperceptibility: It refers to the sensed quality of the screen image in the presence of the water line. As a step of the quality of a watermarked image, the peak signal to resound ratio ( PSNR ) is typically used. PSNR in dBs ( dubnium ) is given below in Eqn. 4
( 5 )
Robustness: It is a step of the unsusceptibility of the water line against efforts to fiddle or degrade it, with different types of digital signal processing onslaughts. In this work, experiments are conducted for hardiness to compaction and resizing.
( a )
( B ) ( degree Celsius )
Fig. 4 ( a ) Watermarked image of ‘head-mri ‘ ( 512×512, bmp ) subjected to JPEG compaction onslaught, at addition k=10 ( B ) Original Watermark SVEC ( 30×20 ) and ( degree Celsius ) Recovered Watermark with JPEG compaction at addition, k=10
Fig. 5 ( a ) Watermarked image of ‘weather-map ‘ ( 512×512, bmp ) subjected to JPEG compaction onslaught, at addition k=10 ( B ) Original Watermark SVEC ( 30×20 ) and ( degree Celsius ) Recovered Watermark with JPEG compaction at addition, k=10
The similarity between the original water line and the extracted water line is measured utilizing the correlativity factor C, is given below in Eqn.6.
( 6 )
N is the figure of pels in the water line, are the original and extracted water line images severally. The correlativity values C will lie between 0 to 1.
5. EXPERIMENTAL RESULTS
Experiments have been conducted on MRI-scan image of HEAD ( 512×152, 8-bit grey graduated table ) . Using the DWT on the image produced four 256×256 bomber sets: LL1, LH1, HL1 and HH1. Since implanting the water line beyond the first DWT degree is more effectual, the water line is embedded in HL2 ( or HH2 ) . The selected 128?128 sub set is divided into 4?4 blocks giving a sum of 1024 blocks. The DCT transform is so applied to each 4×4 block in the chosen bomber set, after which the water line was embedded harmonizing to Eqn. 3 and 4.
Table1 shows the correlativity values between the original water line and the water lines extracted from sub set HH2 after being subjected to different degrees of JPEG compaction. The correlativity values given in Table 2 indicate clearly that the combined DWT-DCT watermarking algorithm is robust against the compaction onslaught. The compaction consequence has besides been tested in the HL2 bomber set and the consequences verified. It was observed that the consequences are better regardless of whether the water line was embedded in HL2 or HH2. Fig.4 and Fig.5 show the extracted water lines from HH2 with assorted degrees of JPEG compaction on themedical image ‘head-mri ‘ and satellite image ‘weather-map’..
The imperceptibility of combined DWT-DCT algorithm is evaluated by mensurating PSNR for the HH2 bomber set. The PSNR values obtained are 46.19, 40.17, 36.65, 34.15 and 32.21 for K values runing from 10 to 50.
In this paper, a combined DWT-DCT digital image watermarking algorithm has been described and tested on the screen image LENA. Watermarking was done by implanting the water line in the 2nd degree DWT bomber set HH2 of the screen image, followed by the application of DCT on the selected DWT sub sets. The combination of the two transforms improved the imperceptibility of the watermarked image by keeping the hardiness to onslaughts.