INTRODUCTION
Invertible data secreting is a highly protected technology to implant clandestine data into host picture indiscernibly and enable the host picture to be regained without any fault upon extraction of the implanted clandestine data which gives additional information to the embedded data. This technology is used in digital watermarking techniques where clandestine data and the image are very important.
Many reversible data hiding methodologies were developed in recent years. Difference between two adjacent pixels is used to embed data in difference expansion method1. A data can be hide using histogram shifting mechanism by Ni et al.2. The lossless compression method is used to create extra space to embed data by Celik et al.3. To improves the performance of reversible data hiding various technologies have been introduced by Luo et al.4, Hong et al.5 and Chang et al.6. Amplitudes of DCT coefficients are used to embed the data7.
In order to increase the information security Zhang8 proposed the lossless data hiding technique in an encrypted image. Zhang9 separable lossless data hiding is introduced. Sorting and prediction method is used for lossless watermarking algorithm10. Ni et al.11 scheme introducing a new embedding mechanism, which increases the data embedding capacity. In this method the image is divided into several numer of sub images and using pathwork theory for each sub image secret data are embeded.
Zeng et al.12 are introduced new algorithm for reversible data hiding using arithmetic difference. The secret can be easily retreived based on the data distribution by Ni et al.11 and Zeng et al.12. De Vleeschouwer et al.13 and Kim et al.14 showed better imperceptibility in their algorithm than Ni et al.11 and Zeng et al.12. Zhao et al.15 introduced a method called multilevel histogram modification. Huang and Chang16 introduced algorithm based on quantized coefficients of Discrete Wavelet Transform (DWT).
MATERIALS AND METHODS
Literature survey: Ramaswamy and Arumugam17 introduced a new lossless data hiding algorithm using adjacency pixel difference. The embedding algorithm steps are given below:
• |
Scan the grayscale image |
• |
Find the pixel difference di for each adjacency pair as follows in Eq. 1: |
• |
Find the peak difference PP |
• |
If di >PP the shift pixel value by 1 as follows in Eq. 2: |
• |
If di = PP, change the pixel value according to the secret bit [0,1] according to Eq. 3: |
In this proposed method the host image is divided in to several 2×2 sub images. Then adjacency pixel difference is calculated for each block. From this difference peak point is calculated. The payload capacity is based on the peak point level. If the peak point is high then payload capacity is also high. This algorithm retrieves the secret without any loss and the image is also recovered with no distortion.
Proposed scheme: An efficient reversible data hiding method is proposed using histogram shifting is given in Fig. 1. Here adjacency pixel difference is used for embedding. Neighbouring pixel in an image is highly correlated. Hence, maximum adjacency pixel difference is nearly zero as shown in Fig. 2a. Let us consider d is the peak point of the histogram and add n with each difference which have value greater than d and this is called histogram shifting, where n value is 3 for 2 bits embedding and for 3 and 4 bits embedding n values are 7 and 15, respectively.
|
Fig. 2(a-b): |
(a) Histogram for absolute difference and (b) Histogram after shifting |
Equation 1 and Fig. 2b show the histogram shifting. Histogram shifting algorithm and embedding algorithm are given in this study. Data retrieving and image reconstruction is reverse process of the data embedding and histogram shifting in Eq. 4:
Histogram shifting
• |
Separate the cover image into several 2×2 sub images as: |
• |
Loop for all sub images |
|
• |
Temp1 = |a11-a21| |
|
• |
Temp2 = |a12-a22| |
|
• |
Difference = |Temp1-Temp2| |
• |
End loop |
• |
Draw the histogram for the difference value |
• |
Let d is the peak point of the histogram |
• |
Loop for all sub images |
|
• |
Temp1 = |a11-a21| |
|
• |
Temp2 = |a12-a22| |
|
• |
If(|Temp1-Temp2|>d) |
|
|
• |
If(|Temp1|>|Temp2|) |
|
|
|
• |
if(a11>a21) |
|
|
|
|
• |
a11 = a11+3 |
Embedding algorithm:
• |
Loop for all sub images |
|
• |
Temp1 = |a11-a21| |
|
• |
Temp2 = |a12-a22| |
|
• |
If(Temp1|-|Temp2|is equal to d) |
|
|
• |
if(secret to be embedded is "00") |
|
|
|
• |
Do nothing |
|
|
• |
Else if(secret to be embedded is "01") |
|
|
|
• |
if(a11>a21) |
|
|
• |
Else if(secret to be embedded is "10") |
|
|
|
• |
if(a11>a21) |
|
|
• |
Else if(secret to be embedded is "11") |
|
|
|
• |
if(a11>a21) |
|
• |
If(|Temp1|-|Temp2|is equal to -d ) |
|
|
• |
if(secret to be embedded is "00") |
|
|
|
• |
Do nothing |
|
|
• |
Else if(secret to be embedded is "01") |
|
|
|
• |
if(a12>a22) |
|
|
• |
Else if(secret to be embedded is "10") |
|
|
|
• |
if(a12>a22) |
|
|
• |
Else if(secret to be embedded is "11") |
|
|
|
• |
if(a12>a22) |
RESULTS AND DISCUSSION
This new lossless data hiding algorithm based on adjacency pixel difference was implemented using jdk1.6. The cover Lena image size 512×512 is given in Fig. 3a. The resultant stego image after embedding is given in Fig. 3b. The recovered image after extracting the secret image is given in Fig. 3c. The corresponding histogram of cover, stego and recovered image of Lena is given in Fig. 3d-f, respectively and observed no difference in histogram level.
The baboon cover and stego images are given in Fig. 4a and b, respectively and recovered baboon image is given in Fig. 4c. Figure 4d-f are shown their corresponding histograms and found to be intact for all the cases original, stego and recovered cover object.
Babara cover, stego and recovered images are given in Fig. 5a-c, respectively and those corresponding histograms are given in Fig. 5d-f, respectively and found to be intact for all the cases original, stego and recovered cover object.
The cameraman cover and stego images are given in Fig. 6a and b, respectively and recovered cameraman image is given in Fig. 6c. Figure 6d-f are shown their corresponding histograms and found to be intact for all the cases original, stego and recovered cover object.
Gold Hill cover and stego images are given in Fig. 7a and b, respectively and recovered Gold Hill image is given in Fig. 7c. Figure 7d-f are shown their corresponding histograms and found to be intact for all the cases original, stego and recovered cover objects.
The man cover and stego images are given in Fig. 8a and b, respectively and recovered man image is given in Fig. 8c. Figure 8d-f are shown their corresponding histograms found to be intact for all the cases original, stego and recovered cover object. The boat cover and stego images are given in Fig. 9a and b, respectively and recovered boat image is given in Fig. 9c.
The butterfly cover and stego images aregiven in Fig. 10a and b, respectively and recovered butterfly image is given in Fig. 10c. Figure 10d-f are shown their corresponding histograms.
|
Fig. 3(a-f): |
(a) Original Lena, (b) Stego Lena, (c) Recovered Lena, (d) Histogram of original Lena, (e) Histogram of stego Lena and (f) Histogram of recovered Lena |
|
Fig. 4(a-f): |
(a) Original baboon, (b) Stego baboon, (c) Recovered baboon, (d) Histogram of original baboon, (e) Histogram of stego baboon and (f) Histogram of recovered baboon |
|
Fig. 5(a-f): |
(a) Original Barbara, (b) Stego Barbara, (c) Recovered Barbara, (d) Histogram of original Barbara, (e) Histogram of stego Barbara and (f) Histogram of recovered Barbara |
The road cover and stego images are given in Fig. 11a and b, respectively and recovered road image is given in Fig. 11c. Figure 11d-f are shown their corresponding histograms.
The manwoman cover and stego images are given in Fig. 12a and b, respectively and recovered manwoman image is given in Fig. 12c.
|
Fig. 6(a-f): |
(a) Original cameraman, (b) Stego cameraman, (c) Recovered cameraman, (d) Histogram of original cameraman, (e) Histogram of stego cameraman and (f) Histogram of recovered cameraman |
|
Fig. 7(a-f): |
(a) Original Gold Hill, (b) Stego Gold Hill, (c) Recovered Gold Hill, (d) Histogram of original Gold Hill, (e) Histogram of stego Gold Hill and (f) Histogram of recovered Gold Hill |
Figure 12d-f are shown their corresponding histograms and found to be intact for all the cases original, stego and recovered cover objects.
This new lossless data hiding algorithm based on adjacency pixel difference method embedding capacity is superior as compared with other method is shown in Table 1. Pepper (512×512) offer good capacity and PSNR. Butterfly host image (512×512) offer low capacity, the stego object offer a favorable PSNR value for steganography and comparable with baboon (512×512)2,17.
|
Fig. 8(a-f): |
(a) Original man, (b) Stego man, (c) Recovered man, (d) Histogram of original man, (e) Histogram of stego man and (f) Histogram of recovered man |
|
Fig. 9(a-c): |
(a) Original boat, (b) Stego boat and (c) Recovered boat |
Table 1: | Hiding capacity for 512×512 grayscale image and image distortion |
|
|
Fig. 10(a-f): |
(a) Original butterfly, (b) Stego butterfly, (c) Recovered butterfly, (d) Histogram of original butterfly, (e) Histogram of stego butterfly and (f) Histogram of recovered butterfly |
|
Fig. 11(a-f): |
(a) Original road, (b) Stego road, (c) Recovered road, (d) Histogram of original road, (e) Histogram of stego road and (f) Histogram of recovered road |
|
Fig. 12(a-f): |
(a) Original manwoman, (b) Stego manwoman, (c) Recovered manwoman, (d) Histogram of original manwoman, (e) Histogram of stego manwoman and (f) Histogram of recovered manwoman |
CONCLUSION
This study gives an overall view about the reversible data hiding technique using histogram modification. This can be achieved by calculating the difference between two neighbouring pixels. Except the edge pixels almost all neighbouring pixels are extremely correlated, this will help to embed huge quantity of secret data when compare to the older ones. For full lossless data retrieving and image recovery is possible to share the peak difference value from the histogram. This technique is successfully compiled and executed using jdk1.8. In the years to come, this powerful algorithm will find a wide range of applications. One such instance is the direction of video sequences where each sequence can be separated into single frames. In each frame this method suggests to apply the same lossless data hiding technique and combine those frames to get the original.
SIGNIFICANCE STATEMENT
• |
Reversible secret data embedding based on histogram modification technique is proposed and executed using jdk1.8 |
• |
Histogram for the absolute pixel difference is calculated and embedding is done only at the peak points |
• |
Full lossless data retrieving and image recovery is achieved |
• |
The PSNR and embedding capacity is improved and comparison is done with the existing methods |