SHORT PAPER International Journal of Recent Trends in Engineering, Vol. 1, No. 1, May 2009
Pioneering the Technique for Invisible Image Watermarking on Color Image Soumik Das1, Pradosh Bandyopadhyay2, Shauvik Paul3, Arindam SinhaRay4 & Dr.Monalisa Banerjee5 Techno India/MCA, Salt Lake, India Email:
[email protected] Email:
[email protected],
[email protected],
[email protected], 5
[email protected] 1] The extracted watermark remains intact. i.e., image authentication is guaranteed. 2] Quality of generated watermarked image is much improved. w.r.t. others. [5] 3] The security issues [3] related to the watermark by using a secret key and a hash function. In current communication, we’ve discussed the technique for embedding the watermark and technique for extraction of the watermark in Section II. We’ve reported experimental analysis in Section III and conclusions are made in Section IV.
Abstract— Accessibility and ease of internet, together with availability of relatively inexpensive digital recording and storage devices has created an environment where duplication, unauthenticated use and maldistribution of digital content has become very easy. This also facilitates unauthorized use, misappropriation, misrepresentation. Thus authentication of multimedia contents (such as image, video, audio) has achieved a sharp attention in recent times. In this perspective we’ve already introduced a framework for invisible image watermarking for color image authentication. In this paper we propose a method for embedding color watermark image to color host image in a more efficient manner. We introduce a new arena of embedding color watermark image at LSB to different positions of color host image in this paper. In proposed watermarking framework, it allows a user with an appropriate secret key and a hash function to verify the authencity, integrity and ownership of an image. We’ve used blind extraction method.
II. TECHNIQUE FOR EMBEDDING WATERMARK AND EXTRACTION OF WATERMARK With contrast to the previous one, a major improvement is being made in this paper. Our already proposed framework will extract two LSB’s of ‘Blue(B)’ value from each pixel of individual tiles of host image using a secret key and a hash function. Instead of embedding the watermark image at two LSB’s of ‘Blue’(B) value of each pixel of individual tiles, here we’ve taken only the LSB of ‘Blue’(B) value of each pixel of individual tiles of host image. As ‘Blue’ is less sensitive to Human Visual System (HVS) [5],[6], we are only manipulating the ‘Blue’ value. The ‘Red’ value and ‘Green’ value of color host image are getting unchanged. The comparative study of the algorithms for 2-bit scheme and 1-bit scheme is reported below.
I. INTRODUCTION
D
igital Watermarking is the process of embedding information into a digital signal.
A watermarking approach provides a persistent link between the authenticator and the content it authenticates. [1] The Digital Image Watermarking (will be referred to as watermarking for rest of the paper) can be classified into two categories - Visible Watermarking and Invisible Watermarking [4], as well as Fragile and Semi fragile watermarking. Our focused area is Invisible Fragile Watermarking. In this kind of watermarking the information is added as digital data to the original, but it can not be perceived as such. Moreover it is highly sensitive to any type of data alteration and tampering can easily be detected [2].
According to our proposed frame work minimum no. of host image block multiplied by no. of bit embedded should be equal to the bit size of RGB. If the size of host image is not sufficient for 1-bit scheme, we can easily apply 2-bit scheme because PSNR value for 2-bit scheme is also very good. Detailed analysis will be found in Section III. Here we’ve made a comparative study between our already proposed framework (2-bit scheme) [8] [9] and the new framework (1-bit scheme).
At best to our knowledge we are pioneering the technology which can embed a color watermark image in a color host image. In this sort of watermark, our proposal with 2-bit LSB’s scheme (will be referred to as 2-bit scheme for rest of the paper) is already accepted [8]. In this paper we’ve improved our scheme to 1-bit LSB scheme (will be referred to as 1-bit scheme for rest of the paper) to achieve a much better result regarding PSNR value (Detailed Analysis is reported in the sub section B of Section III). Our proposed framework will ensure-
A. Algorithm for Embedding Watermark using 1-bit Scheme: Let the color host image be ‘Y’. A color watermark ‘A’ will be inserted in it and again will be extracted from it for authentication. The color host image is divided into twenty-four equal blocks (Yi), where each block size is same with color watermark ‘A’. In the following discussion, we concentrate on inserting and extracting some bit information of color watermark ‘A’ into the 508
© 2009 ACADEMY PUBLISHER
SHORT PAPER International Journal of Recent Trends in Engineering, Vol. 1, No. 1, May 2009 corresponding color host image block Yi and form The binary information of blue value of a pixel of A is watermark embedded image block Wi. given by Ab. [insert_color_watermark_1bit] Input:
Color host image block (Yi), Watermark image (A), Secret key (K). Output: Watermark embedded image block (Wi).
Step 2: The LSB of the binary values of Yip is converted to zero to form Li.
Step 1: Yip= get_pixel_info (Yi ). Step 2: Li = onebit_to_zero (Yip ). Step 3: Hi = hash_to_generate_mesg_digest (Li , K). Step 4: Ai = extract_onebit (A). Step 5: Xi = compute_XOR (Hi, Ai) Step 6: Wi = replace (Yip , Xi ). B. Function Definitions: get_pixel_info (Yi): This function will bring the binary information of blue value of a pixel of host image block (Yi) that forms Yip . onebit_to_zero (Yip): This function will convert the LSB of the binary values of Yip to zero to form Li. hash_to_generate_mesg_digest (Li, K): This function will generate the message digest (Hi) depending on the value of Li and a 1024-bit secret key K. extract_onebit (A): This function will extract LSB of each pixel of watermark image (A) that forms Ai. compute_XOR (Hi, Ai): This function will perform XOR operation between Hi and Ai to form Xi. replace (Yip , Xi ): This function will replace the LSB of Yip with the value Xi which we got by XORing Hi and Ai to form Wi.
Step 3: Hi is generated depending on the value of Li and The 16-bit secret key (K).
Step 4: Ai formed by extracting LSB of blue value of each pixel of A.
Step 5: Xi is formed by XORing Hi and Ai.
C. Execution of Algorithm for Embedding Watermark using 1-bit Scheme: For example we are taking a block (Yi) of size 4X4 of image CHILD.jpg as our host image and LOGO1.png as our watermark image (A). 16-bit secret key (k) is also defined. Now we are stepping forward as per our proposed algorithm for embedding watermark.
Step 6: Wi is formed by replacing the LSB of Yip with Xi.
Input: Host Image block
Fig.1 shows the schematic diagram for embedding watermark to the host image. Where Yi is referred as host image block, A is referred as watermark image; Wi is referred as watermarked image block. Fig.2 shows how the prescribed technique is applied to the host image CHILD.jpg.
Watermark Image:
Secret Key: K= 1111 0000 1010 1100 Step 1: The binary information of blue value of a pixel of Yi is given by Yip.
509 © 2009 ACADEMY PUBLISHER
SHORT PAPER International Journal of Recent Trends in Engineering, Vol. 1, No. 1, May 2009
K=
Secret Key: 1111 0000 1010 1100
Step 1: The LSB of the binary values of Wi is converted to zero to form L’i.
Step 2: Hi is generated depending on the value of L’i and The 32-bit secret key (K). Fig. 1 Schematic Diagram for Embedding Watermark
Step 3: Ri formed by extracting LSB of blue value of each pixel of Wi.
A
Y +
Step 4: A’i is formed by XORing Hi and Ri.
Observation: We can easily see that A’i= Ai. So our watermark image is retained intact after extraction. Fig.3 shows the schematic diagram for extracting watermark from watermarked image. Where Wi is referred as watermarked image block, A’i is referred as extracted watermark image block.
W Fig. 2 Scheme Applied on CHILD.jpg
D. Algorithm for Extracting Watermark using 1-bit Scheme: [extract_color_watermark_1bit] Input: Watermark embedded image block (Wi), Secret key (K). Output: Extracted watermark image block (A i). Step 1: Step 2: Step 3: Step 4:
L i = onebit_to_zero (Wi). H i = hash_to_generate_mesg_digest (L i , K). Ri = retrieve (Wi). A i = compute_XOR (H i ,Ri)
Function Definition(Extended): retrieve(Wi): This function will retrieve LSB of blue value of each pixel of watermarked image (Wi) that forms Ri. Other function are already defined during embedding. E. Execution of Algorithm for Extracting Watermark using 1-bit Scheme:
Fig. 3 Schematic Diagram for Extracting Watermark
We’ve applied our proposed framework on several images. The result sets are reported in the next section. Input: Watermarked Image Block
III. EXPERIMENTAL ANALYSIS
510 © 2009 ACADEMY PUBLISHER
SHORT PAPER International Journal of Recent Trends in Engineering, Vol. 1, No. 1, May 2009 ensures that the extracted watermark retained as it is, Theoretical excellency of our proposed scheme has since it attains maximum value of nc, i.e., unity. merged with our detailed experimental analysis. We’ve tested our proposed framework on thousand different TABLE II color images. And we’ve found a set of excellent results NC VALUE ANALYSIS of our experiments. In this section we’ve concentrated on Image Name Size nc value three major aspects: Californian_Fruit_Salad 290 * 260 1.00 A) Persistency of Quality after Embedding and Extracting Animal.jpg 1024* 768 1.00 the Watermark / Invisibility to HVS. Storfe_kuer.jpg 1500 * 1000 1.00 B) Accurate Extraction. C) Security Issues. C. Security Issues: Security issue is assured with a secret key and a hash function. Detailed analysis will be found in [4].
A. Persistency of Quality after Embedding and Extracting the Watermark / Invisibility to HVS: Our proposed framework has been tested on different color images. One of the following images which ensures the persistency of their quality.
CONCLUSION The detailed experiment has been performed and it is found that our proposed framework is able to embed the color watermark images to color host images and perceptually the watermark is not visible in the watermarked image. We’ve used blind method for watermark extraction. With addition to that we also ensure that the extracted watermark remains intact. Security issue is assured with a secret key and a hash function.
1. Tested on Californian_Fruit_Salad. jpg
a) Host Image
b) Watermark
REFERENCES [1]
[2] c) Watermarked Image
d) Extracted Watermark
From the above image it is clear that the watermark is entirely invisible to the HVS. In this regard PSNR [7] is most commonly used as a measure of quality in reconstruction of image. 3rd column of Table. I shows the analysis on PSNR values for three different color host images using 1-bit scheme and 4th column for 2-bit scheme. TABLE I PSNR VALUE ANALYSIS USING 1-BIT & 2-BIT SCHEME Image Image PSNR PSNR Name Size value value (1-bit) (2-bit) Californian_Fruit_ Salad.jpg 290 * 260 81.04 64.88 Animal.jpg 1024* 768 89.03 75.16 Storfe_kuer.jpg 1500 * 1000 90.94 77.84
[3]
[4]
[5]
[6] [7] [8]
Observation: We can easily see that the PSNR value is getting much better in 1-bit scheme than our previously proposed 2-bit scheme.
[9]
B. Accurate Extraction: When we are going to extract the watermark from watermarked image, the watermark is retained as it is. Table. II is provided in this regard. The quantitive similarity measurement between the referenced watermark and extracted water mark is computed by normalized correlation (nc). 3rd column of Table.II shows the nc values for three different color host images It 511 © 2009 ACADEMY PUBLISHER
Christine I. Podilchuk, “Image-Adaptive Watermarking Using Visual Models.”, Proceedings of IEEE Journal on Selected Areas in Communications, Vol. 16, no. 4, May 1998. Oktay Altun, Gaurav Sharma, Mehmet U. Celik, and Mark F. Bocko, “A Set Theoretic Framework for Watermarking and Its Application to Semifragile Tamper Detection”, Proceedings of IEEE Transactions on Information Forensics and Security, Vol. 1, no. 4, December 2006. R. Wolfgang and E. J. Delp, “A watermark for digital images,” Proceedings of IEEE International Conference on Image Processing, Vol. 3, pp.219-222, 1996. M. Banerjee, “Theory and application of cellular automata for authentication and watermarking”, PhD thesis, Jadavpur University, 2007. Chang-Tsun Li and Yue Li, “Random Index Modulation Based Fragile Watermarking Scheme For Authenticating Colour Images”, Proceedings of International Conference on Intelligent Information Hiding and Multimedia Signal Processing, 2008. Ranjan Parek,.”Principle of Multimedia”, Tata McGraw Hill, New Delhi, 2006, ISBN: 0-07-058833-3. http://en.wikipedia.org/wiki/Peak_signal-to-noise_ratio. Soumik Das, Pradosh Bandyopadhyay, Shauvik Paul, Arindam Sinha Ray and Dr. M. Banerjee, “Artifacts On Classical Theory Of Invisible Image Watermarking On Color Image,” Published in MACMILLAN Advanced Research Series (ISBN 023-063-759-0) Page No. 248, 1st International Conference on Computer, Communication, Control and Information Technology, India, 2009. Soumik Das, Pradosh Bandyopadhyay, Shauvik Paul, Arindam Sinha Ray and Dr. M. Banerjee, “A New Introduction towards Invisible Image Watermarking on Color Image,” Accepted in IEEE International Advance Computing Conference (IACC’09), India, 2009.
