KR20130085264A - Integrity checking process of image using reversible watermarking process tecjnique - Google Patents

Abstract

PURPOSE: An integrity authentication method of an image which uses a reversible watermarking technology is provided to authenticate integrity without damage to picture quality by applying a watermarking algorithm to a JPEG image compression format. CONSTITUTION: A quantization DCT coefficient is restored by extracting an authentication code from the restored watermarking quantization DCT coefficient. An original image is restored through the inverse quantization of the restored quantization DCT coefficient, inverse DCT, and color space variation. The authentication code is regenerated from the restored original image. The integrity of the image is confirmed by comparing the extracted authentication code with the regenerated authentication code. [Reference numerals] (AA) Compression process with loss; (BB) Image divided into a 16 x 16 block unit; (CC) Convert into an YCbCr to an RGB color domain; (DD) Divide an 8 x 8 block unit, Descrete cosine convert; (EE) DCT coefficient quantization; (FF) Restored RGB; (GG) Restore YCbCr to RGB; (HH) Inverse DCT; (II) Inverse quantization; (JJ) Quantization DCT coefficient; (KK) Down-sample a 16 x 16 block into an 8 x 8; (LL) Gray scale each channel with the average value; (MM) 8 x 8 feature block; (NN) Select 8 coefficients and insert into the 16 x 16 block; (OO) Quantization; (PP) Feature block DCT coefficient; (QQ) Water mark inserted quantization DCT coefficient; (RR) Water mark inserting process; (SS) Water mark inserted JPEG image; (TT) Entropy coding; (UU) Compression process without loss

Description

Integrity checking Process of Image using Reversible Watermarking Process Tecjnique}

The present invention relates to a method for authenticating an image integrity, and more particularly, by applying a watermarking algorithm to a JPEG image compression format to authenticate integrity without compromising image quality, an image of an image that can simultaneously preserve reliability and practicality of image content. It relates to a method for authenticating integrity.

With the development of hardware and software, the production and distribution of multimedia contents such as moving pictures, photos, music, and publication documents have been expanded, and the market size is growing every year with the introduction of related services. Multimedia contents are easy to copy and modify contents due to the characteristics of digital data that are electronically stored and distributed, and easily cause forgery and illegal distribution problems. Therefore, as the content market grows in size and the penetration of high-performance hardware capable of manipulating multimedia data is accelerating, the importance of security technologies applicable to multimedia contents is increasing day by day.

In particular, in many fields such as medical, military, and art, reliability and security of multimedia data are required as important criteria. In the case of surveillance camera images or black box images of vehicles, manipulation may cause human and material damage or lose its function as legal evidence. Since medical imaging is a field that can be directly connected to a patient's health or life, these imaging data need high reliability.

Digital watermarking algorithms are classified into two types: robust watermarking algorithms and fragile watermarking algorithms. Most of the robust watermarking algorithms are designed to protect the watermark data as much as possible even if the image is attacked due to image distortion or distortion, so it is suitable for the watermarking field for data hiding or preservation. The soft watermarking algorithm is suitable for the watermarking algorithm used for video authentication or integrity assurance. This is because the soft watermarking algorithm is useful for authenticating the original image because the watermark data is easily damaged by an attack.

Reversible watermarking is one of flexible watermarking techniques. It is possible to partially or completely reconstruct the original image by removing the watermark inserted in the image. The reversible watermarking techniques studied for conventional video authentication are mainly a method of inserting and detecting a watermark for an image that has not been compressed. This method has a disadvantage in that watermark is easily damaged during image compression. It is difficult to apply to an image format for compressing and storing the image, which causes a problem of poor practicality.

In the prior art related to image authentication technology, References [1] and [2] describe a method of inserting data into the LSB as a method for recording authentication data or patient information for verifying originality in medical images. Suggested. The disadvantage of LSB-based watermarking algorithms is their weak security. This is because if the LSB value of the pixel data is simply replaced with watermark data, forgery can be easily performed by simply copying the LSB data to another forged image.

Ref. [3] proposed DWT-based multiple watermarking technology for the purpose of authentication, distortion detection, and copyright protection in digital images. Watermark information was inserted in the LSB of the LL band, which is a low frequency region, to detect the deformation position of the image, and watermark information was inserted by exchanging wavelet coefficients of the high frequency portion for copyright protection.

[4] proposed a hierarchical data concealment technique for detecting and restoring image forgery. However, since these methods are irreversible watermarking algorithms that cannot recover the original after inserting the watermark, they are not used in the field where image quality is important.

Ref. [5] proposed a method of inserting a watermark by adjusting the coefficient of the high frequency region of the DCT conversion region by using the histogram correction of the image, and the loss of the entire image through the adjustment of the brightness histogram during the insertion process. This happens, it is difficult to delete the inserted watermark and restore the original has a disadvantage that the reversibility of the watermarking algorithm.

Because multimedia content inevitably has a large amount of data, it includes processing such as JPEG compression. Such processing is regarded as an attack in terms of watermark, and studies on image authentication techniques that have secured robustness against these attacks Is continued in References [6] to [8] and the like.

Reversible watermarking is mainly used as a watermarking technique for image content, in which image quality is an important issue. Reversible watermarking technology can restore the original image quality of the content partly or completely after going through the process of authenticating the content or extracting the hidden information. Studies on reversible watermarking insert messages using different image features to ensure perceptual transparency and complete reversibility.

In this regard, Ref. [9] inserts a message in the empty space after compressing the bit plane using the lossless compression technique, and Ref. [10] inserts data into the transform coefficients in the frequency domain.

Ref. [11] proposed a reversible watermarking algorithm that inserts a watermark into the data during lossless processing during JPEG compression, and restricts the insertion area into the quantization table in order to reduce the quality degradation caused by the insertion process. The method of inserting was used. However, a method of embedding a watermark on a single channel of a gray image is difficult to apply to a color image as it is.

Ref. [12] proposed a reversible watermarking algorithm that inserts data using a progressive histogram of an image. However, the method using the difference value histogram requires additional data concealment or separate transmission as an alternative to the overflow problem that occurs during insertion.

On the other hand, the general JPEG compression algorithm, as shown in Figure 1, is largely composed of two stages of a lossy compression process and a lossless compression process. The lossy compression process consists of color space transform, discrete cosine transform, quantization, etc., and data loss occurs. The lossless compression process consists of an entropy encoding step including run-length encoding and huffman coding, and there is no data loss.

Specifically, in the lossy compression step, RGB data of each pixel is first converted into YCbCr data. The YCbCr color space is similar to the color space used for NTSC or PAL color television transmission. The Y component has pixel luminance information, and the Cb and Cr components have color information. Because the human eye is more sensitive to luminance components than color components, it converts to the YCbCr color space to compress more color information. Real number operation occurs during color space conversion, so the information is not completely preserved. Therefore, even if it does not go through the subsequent quantization process, it does not completely match the original. In the JPEG compression algorithm, there is a process of increasing the compression rate by reducing the number of Cb and Cr components through the chroma subsampling process. In this process, it is difficult to preserve the storage capacity and the image quality of the watermark embedded image.

After color space transformation, the entire image is divided into 8x8 block units, and each block is transformed into frequency space using a two-dimensional discrete cosine transform (DCT). The human eye is indistinguishable at high frequency brightness changes. Based on this fact, a large part of the high frequency component information can be discarded by dividing each component in the frequency domain into a specific quantization table as shown in FIG. 2 and taking only integer quotients. The most data loss occurs at this stage of the entire JPEG compression stage. After this step, most of the high frequency components are either zero or close to zero or positive.

In the lossless compression step, the zig-zag alignment is performed on the lattice components of the 8x8 DCT block, which is the basic unit of JPEG compression, in order from low frequency components to high frequency components, and run-length encoding is performed for the repeated zeros. Apply, and perform huffman coding on the result. Instead of huffman coding, arithmetic coding is also possible.

Prior Art References

[1] X. Q. Zhou, H. K. Huang, S. L. Lou, "Authenticity and integrity of digital mammography images," IEEE Trans. Medical Imaging, Vol. 20, No. 8, pp. 784-791, 2001.

[2] A. U. Rajendra, D. Anand, B. P. Subbanna, U. C. Niranjan, "Compact storage of medical images with patient information," IEEE Trans. Information Technology in Biomedicine, Vol. 5, No. 4, pp. 320-323, 2001.

[3] Kang Tae-hwan, Jang Ho-hyun, Kim Dong-seo, Nak-geun Joo, "DWT-based Multiple Water Marking for Detection of Image Indentation and Deformation", Proceedings of the Korean Society of Contents Conference, Vol. 3, No. 2, pp. 480-484, 2005.

[4] P. L. Lin, C. K. Hsieh, P. W. Huang, "A hierarchical digital watermarking method for image tamper detection and recovery," Pattern Recognition, Vol. 38, No. 12, pp. 2519-2529, 2005.

[5] Hyo-Chul Kim, "High Reliability Image Authentication Using Histogram Correction," Journal of Korea Multimedia Society, Vol. 13, No. 7, pp. 1088-1094, 2010.

[6] A. Shiozaki, M. Iwata, A. Ogihara, "Reversible Watermarking Method for JPEG Image," IEICE Trans. on Information and Systems, Vol. E91-D, No. 7, pp. 2068-2071, 2008.

[7] X. Shi, F. Liu, D. Gong, and J. Jing, “An Authentication Watermark Algorithm for JPEG images,” Proc. Of the International Conference on Availability, Reliability and Security, pp. 584-588, 2009 .

[8] X. Wang, J. Wang, and H. Peng, "A Semi-fragile Image Watermarking Resisting to JPEG Compression," Proc. of the International Conference on Management of e-Commerce and e-Government, pp. 498-502, 2009.

[9] M. U. Celik, G. Sharma, A. M. Tekalp and E. Saber, "Lossless generalized-LSB data embedding," IEEE Trans. on Image Processing, Vol. 14, No. 2, pp. 253-266, 2005.

[10] S. Lee, C. D. Yoo and T. Kalker, "Reversible image watermarking based on integer-to-integer wavelet transform," IEEE Trans. on Information Forensics and Security, Vol. 2, No. 3, pp. 321-330, 2007.

[11] Haknam Choi, Jongwon Kim, Jongwook Choi, Hakil Kim, "Reversible Watermarking in JPEG Compression Domain," Journal of Information Processing Society, Vol. 17, pp. 121-130, 2007.

[12] Dong-Kyu Yeo, Hae-Yeon Lee, "Block-based Image Authentication Algorithm using Reversible Watermarking based on Difference Values", Journal of Information Processing Society, 2011 (in press).

The present invention provides a method of inserting a specific watermark in the quantized data during JPEG compression process, assuming that the damage during image restoration occurring in the final stage of the modification and forgery attack is the main ground of the attack, and the deformation of the image is assumed. By detecting, it is an object to provide a method for authenticating the integrity of an image.

Integrity authentication method of the image using the reversible watermarking technology of the present invention for achieving the above object, (a) color space conversion, discrete cosine transform (hereinafter referred to as 'DCT') and quantization process for the original image Extracting quantized DCT coefficients through; (b) generating an authentication code using the extracted quantized DCT coefficients, and inserting the generated authentication code into the quantized DCT coefficients extracted in a lossy compression process to watermark; (c) entropy coding the watermarked quantized DCT coefficients to generate a compressed image; (d) entropy decoding the compressed image to reconstruct the watermarked quantized DCT coefficients; (e) restoring the quantized DCT coefficients by extracting an authentication code from the restored watermarked quantized DCT coefficients; (f) reconstructing the original image through inverse quantization, inverse DCT, and color space change of the reconstructed quantized DCT coefficients; (g) regenerating an authentication code from the restored original image; And (h) comparing the authentication code extracted in step (e) with the authentication code reproduced in step (g) to confirm the integrity of the image.

Here, the generation of the authentication code of step (b) is characterized in that to obtain the quantized DCT coefficients through the step (a), and to generate and generate the n highest values in the order of zig-zag from the obtained quantized DCT coefficients And, inserting the authentication code of step (b), shifting the positive values of the quantized DCT coefficients, characterized in that to insert the authentication code between 0 and the shifted positive value, the authentication code of step (g) The reproducibility is characterized by extracting and generating quantized DCT coefficients by DCT transforming and quantizing the original image reconstructed in step (f).

In addition, in the present invention, a series of the above process, the process of converting the original image of the upper block (N x N block) to the unit image of the lower block (N / 2 x N / 2 block), and then compressing and restoring This is characterized in that it is made.

The watermarking algorithm of the present invention can be inserted and utilized in the JPEG compression process based on the JPEG image compression standard, so high utilization is expected.

It is also sensitive to post-modification attacks, which makes it possible to reliably determine the originality and integrity of an image.

In addition, by using the DCT coefficient of the sampled image block as an authentication code, partial restoration is possible when the image is damaged according to the authentication code insertion position. By removing the inserted watermark, the image quality of the compressed JPEG image can be restored without inserting the watermark.

1 is a view showing a general compression process of a JPEG file,
2 is a diagram illustrating a standard quantization table used in a JPEG compression process;
3 is a view showing a compression process of a JPEG file according to an embodiment of the present invention;
4 is a view showing the authentication code generation and insertion process according to an embodiment of the present invention,
5 is a diagram illustrating a watermark embedding process through histogram shifting according to an embodiment of the present invention;
6 is a view showing an image restoration process through image authentication and watermark removal according to an embodiment of the present invention;
7 is a diagram illustrating a histogram restoration process according to an embodiment of the present invention;
8 is a diagram illustrating a JPEG compressed image according to an embodiment of the present invention; and
FIG. 9 is a diagram illustrating a difference between a watermark embedded image and a non-inserted image of the JPEG compressed image of FIG. 8.

JPEG compression algorithm is one of the most popular image compression algorithms in the world. The present invention proposes a watermarking technique applicable to an image compression algorithm conforming to the JPEG standard. The watermarking technique of the present invention is a soft watermarking algorithm, which is susceptible to correction of an image, thereby easily damaging the watermark. Therefore, it is possible to detect whether the image is deformed when an attack such as a modification to the image occurs. In the JPEG compression standard, the compression rate is determined by the multiple of the quantization table used for quantization of DCT coefficients. Therefore, the water inserted into the quantized DCT coefficient data is restored when the image is partially modified or restored after the compression ratio is changed. The mark data is damaged. In the present invention, it is assumed that the damage during image restoration occurring at the final stage of the modification and forgery attack is the main basis of the attack, and the deformation of the image is detected.

JPEG files generated by the algorithm of the present invention are compatible with the JPEG compression format, and there is no significant degradation or deterioration of the image quality when the watermark is inserted. Therefore, a JPEG image generated through the algorithm of the present invention can check or output an image with a general image viewer.

In addition, the JPEG compression algorithm of the present invention undergoes a lossless compression process after a lossy compression process. In other words, the data after the lossy compression process that causes data loss in the encoding process is preserved through the lossless compression process and the entropy decoding process of the decoding process. Therefore, the watermark embedding method is designed by inserting a specific watermark into the quantized data in the JPEG compression process and preserving the inserted watermark simultaneously with image compression.

BRIEF DESCRIPTION OF THE DRAWINGS The above and other objects, features and advantages of the present invention will be more apparent from the following detailed description taken in conjunction with the accompanying drawings, in which: FIG. Hereinafter, referring to the accompanying drawings, a process of compressing a JPEG image file while generating an authentication code and inserting a watermark, and a process of extracting a watermark and authenticating a restored image while restoring the JPEG image file will be described. do.

[Generate and insert verification code]

3 is a view showing a compression process of a JPEG file according to an embodiment of the present invention, Figure 4 is a view showing a process of generating and inserting the authentication code according to an embodiment of the present invention, Figure 5 is an embodiment of the present invention The figure shows a process of embedding a watermark through histogram shifting.

The basic block size of the JPEG compression method is 8x8 pixels, and the image authentication method according to the embodiment of the present invention is formed in blocks of 16x16 pixels. In order to increase the authentication rate and minimize the probability of false positive detection, a sufficient length of authentication code should be used. When using a 64-bit length authentication code, 2 ⁶⁴ types of authentication codes can be generated. Therefore, the probability of false detection by chance is 1 / 2 ⁶⁴ , or 5.42101E-20, rarely occurs in reality. However, the experiment confirmed that the degradation of image quality is severe when inserting a 64-bit authentication code into an 8x8 block. Therefore, in the present invention, after generating an authentication code in units of 16x16 blocks, 16 bit bits are divided and inserted into four 8x8 subblocks.

The entire process of generating the authentication code and inserting the watermark is shown in FIG. As shown, since both DCT and quantized DCT block data are lost in the previous step, an authentication code is inserted based on the histogram of the quantized DCT block (hereinafter, referred to as a QDCT block) data. The authentication code is generated through DCT conversion on the 16x16 pixel unit in lossy compression. The authentication code generated by using the DCT transformation may be used to partially restore the lost image when a portion of the image is lost. Authentication code generation method is applied to the authentication code generation method of Ref. [12].

First, the process of generating an authentication code in detail, the authentication code downsamples a 16x16 block into an 8x8 gray image block, generates a DCT coefficient block through a discrete cosine transform, and then quantizes it to obtain a quantized DCT coefficient. For this, the top eight values in zig-zag order are used as 64-bit authentication codes. The quantization table used for the DCT coefficient block quantization for generating the authentication code uses the luminance quantization table used for the DCT coefficient quantization of the Y channel in the JPEG compression step. If part of the image is compromised by an attack, the authentication code can be used to recover some of the damage if the authentication code is preserved in another block.

In detail, the process of inserting the authentication code, the authentication code is generated for 16x16 blocks, JPEG image compression is performed in units of 8x8 blocks. Therefore, four blocks of 8x8 pixels are merged and used as one authentication unit. The 64-bit authentication code data is divided into four 16-bit data and inserted into four 8x8 blocks, each of which constitutes a 16x16 block. The authentication code data insertion method is inserted into the quantized QDCT block data by histogram shifting. An authentication code insertion method using authentication code division and histogram shifting is shown in FIGS. 4 and 5.

Specifically, there is a sufficient number of zeros in the QDCT block data to insert an authentication code. After shifting the positive values of the coefficient values constituting the QDCT block data, the watermark bit pattern is inserted in place of the 0s and 1s. 8, 4, and 4 bit patterns are inserted into the QDCT block data of the Y, Cb, and Cr channels, respectively, in consideration of the effect of the constants constituting the quantization table on the deterioration of image quality during the quantization process. In order to minimize image quality deterioration due to shifting during insertion, only positive values up to the position where the authentication code data is inserted are shifted. Insertion of authentication code follows the following rules.

1) Insert QDCT coefficients in zig-zag order.

2) DC components are excluded from insertion and shifting.

3) Shift the positive value to the right, modify the zero value to insert the watermark bit pattern, and leave the negative value as it is.

4) Positive values after insertion is not shifted.

QDCT coefficient data has less effect on image quality as it comes to the front of zig-zag order because of quantization table characteristics. Therefore, the authentication code data is inserted from the front to the rear in the zig-zag order at the time of insertion. Since the DC component has a large influence on image quality deterioration and there is little possibility that zero occurs, shifting or insertion is performed except for the DC component. The authentication code data is inserted in the 0 position of the QDCT coefficient data in the form of bit patterns of 0 and 1, and shifts only the positive number found until the end of insertion, and the coefficient values of the non-inserted position are left as they are. Therefore, only the 'positive component from the original QDCT value to the point where the watermark insertion is left' and 'the value of the position where 1 of the watermark insertion point is inserted' are changed, so the impact of the authentication code insertion process on the image quality is minimal. .

This authentication code insertion process is performed according to Equation 1. ZQDCT is QDCT arranged in zig-zag order.

[Equation 1]

In the JPEG image compression process, QDCT block data generates coefficient values that do not exceed a maximum of 255 at the general compression ratio. In addition, since QDCT block data allows data exceeding 8 bits based on the number of storage bits when it is saved to a file after entropy encoding, overflow problem that occurs in general image watermark embedding algorithm does not occur. Therefore, no storage or delivery of additional information, such as an overflow map, is necessary.

[Watermark Extraction, Image Restoration and Authentication]

FIG. 6 is a diagram illustrating an image restoration process through image authentication and watermark removal according to an embodiment of the present invention, and FIG. 7 is a diagram illustrating a histogram restoration process according to an embodiment of the present invention.

The purpose of the algorithm of the present invention is to guarantee the originality of the image in the JPEG format, and when the image of the JPEG format is resaved, the existing image is changed due to the difference in the compression ratio according to the quantization table or the program used in the compression process according to the modified program. The DCT coefficient data of is changed so that authentication is not possible. This image resave process must go through any attack scenario, so the originality of the image can be considered important evidence. Since the watermark inserted in the algorithm of the present invention is damaged over the entire part of the image at the time of re-storing, it is possible to sensitively determine whether the content is tampered with or forged.

The image verification process uses the JPEG decoding process as it is, but extracts the verification code, regenerates the verification code for comparison, and removes the verification code from the QDCT block data to bring the image quality to the level of a general JPEG encoded image. The part to restore is added.

The JPEG decoding process is performed in the reverse order of the encoding process, and the entropy decoding process, which is a restoration of lossless compression, can acquire 8x8 pixel QDCT block data having a watermark. The authentication code data extracts a total of 16 bits of data from 8 bits on the Y channel and 4 bits on the Cb and Cr channels. The 64-bit authentication code is restored by combining the 16-bit data extracted from four 8x8 pixel blocks forming the 16x16 pixel authentication block. The image authentication and image quality restoration process is shown in FIG. The detection of the authentication code is extracted to the quantization coefficient value as shown in Equation 2.

[Equation 2]

To verify the originality of the image, the verification code for the comparison is regenerated through the verification code generation process in the same way as the authentication code generation method in the encoding process. The image authentication is performed by comparing the reconstructed authentication code with the extracted authentication code through DCT conversion, quantization, etc. on the 16x16 block image reconstructed into an RGB image.

The QDCT coefficient data with the authentication code inserted can restore the original data by reverse shifting positive values on the histogram. Therefore, at the same time as the process of extracting the inserted watermark, an image without the watermark inserted can be restored. Since the verification code generated during the watermark insertion process is generated based on the encoded image rather than being inserted, the verification code can be generated by the same process as the insertion using the restored image, and compared with the extracted verification code. Authentication can be performed. Normally, the encoded image with the watermark inserted must be exactly the same as the extracted authentication code. However, if the image is changed through an attack such as a modification or a compression ratio change, the extracted authentication code and the reproduced authentication code are different, and thus, the attack of the image can be determined.

The algorithm in the present invention is developed on the basis of reversible watermarking to remove the authentication code from the JPEG image embedded with the watermark and restore the quality of the compressed JPEG image without the watermark inserted. Investigating the remaining components except for the highest DC component in the QDCT block data in the order of zig-zag in order, inverting the positive numbers that appear until the inserted bit string data is extracted as many inserts per channel as shown in FIG. 7. Shift to restore the QDCT block data before watermark insertion. The QDCT block data reconstructed in the inverse quantization process and inverse discrete cosine transform after the entropy decoding and authentication step is used to restore the quality of the encoded JPEG image without the watermark inserted. Image reconstruction is performed according to Equation 3.

[Equation 3]

In the following, we look at the performance of the algorithm of the present invention. FIG. 8 is a diagram illustrating a JPEG compressed image according to an embodiment of the present invention, and FIG. 9 is a diagram illustrating a difference between a watermark embedded image and an uninserted image of the JPEG compressed image of FIG. 8.

In order to test the performance of the proposed algorithm, eight 8-bit color 512x512 size images from the University of Southern California-Signal & Image Processing Institute (USC-SIPI) image database were used, and an image compressed in JPEG format is shown in FIG. 8. The difference image between the watermark embedded image and the non-inserted image is shown in FIG. 9. The quality of the compressed and watermark embedded images was measured by PSNR (dB) calculated by Equations 4 and 5.

[Equation 4]

[Equation 5]

Here, the variables M and N represent the width and height of the image, respectively, p (i, j) represents the original image, and p '(i, j) represents the degraded image. Watermarked and non-embedded images actually show a very small difference, and image quality deterioration is hard to see with the naked eye. In the difference image shown in FIG. 9, the brightness of the noise pixel was adjusted to confirm the location of the noise. In addition, reversible watermarking algorithms can be used to restore the compressed image quality, typically using the JPEG algorithm. Table 1 shows the differences in image quality between JPEG compression and watermark embedding of images.

PSNR comparison between watermark embedded and non-inserted compressed images (unit: dB) image Embedded video Uninserted image Difference Airplane 34.33 37.68 3.35 Baboon 30.31 31.64 1.33 House 32.95 35.66 2.71 Lena 33.75 36.34 2.59 Pepper 33.02 35.16 2.14 Sailboat 31.82 33.55 1.73 Splash 35.18 38.42 3.24 Tiffany 33.68 36.13 2.45 Average 33.13 35.57 2.44

As can be seen from Table 1, the image quality in JPEG compression of an uninserted image is maintained at 80-85% of the original quality. 16-bit data per 8x8 pixel is inserted in the 8-bit 3-channel color image for the inserted image, and more data can be inserted, but the quality of the inserted image is degraded. The deterioration of the quality of the compressed video occurring at the time of embedding the watermark is not significantly reduced to 2.44 dB on average based on the PSNR.

The algorithm of the present invention is a watermark embedding method for use in a JPEG image compression method, and the image compression ratio after inserting watermark data may also be a measure of performance evaluation. The compression ratio was compared with the capacity including the header information of the bitmap format file of the original image and the converted JPEG format file.

Comparison of Compression Ratio of Original and Uninserted Compressed Watermarks image Embedded video Uninserted image Difference Airplane 92.27% 93.77% 1.50% Baboon 84.79% 86.69% 1.90% House 90.26% 91.84% 1.58% Lena 92.11% 93.77% 1.66% Pepper 91.18% 92.90% 1.72% Sailboat 89.37% 91.07% 1.70% Splash 93.19% 94.59% 1.40% Tiffany 92.00% 93.55% 1.55% Average 90.65% 92.27% 1.63%

As shown in Table 2, the JPEG compression rate of the raw image is 92.27% of the original compression ratio. In addition, according to an embodiment of the present invention, the watermark-inserted image shows an average compression ratio of 90.65%, indicating that the compression performance is not significantly reduced. Since the number of zeros included in the quantized DCT coefficient data is reduced during the watermark embedding process, the compression performance may also decrease as the capacity of the embedded data increases due to the feature of run-length encoding coding during the lossless compression process.

As described above, the watermarking algorithm according to the embodiment of the present invention can be inserted and utilized in the JPEG compression process based on the JPEG image compression standard, and thus high utilization is expected. It is also susceptible to post-modification replay attacks, which makes it possible to reliably determine the originality and integrity of an image. Also, by using the DCT coefficient of the sampled image block as an authentication code, partial restoration is possible in case of image damage according to the insertion position of the authentication code. By removing the inserted watermark, the image quality of the compressed JPEG image can be restored without inserting the watermark.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments.

Claims (5)

(a) extracting quantized DCT coefficients through color space transform, discrete cosine transform (hereinafter, referred to as 'DCT'), and quantization for the original image;
(b) generating an authentication code using the extracted quantized DCT coefficients, and inserting the generated authentication code into the quantized DCT coefficients extracted in a lossy compression process to watermark;
(c) entropy coding the watermarked quantized DCT coefficients to generate a compressed image;
(d) entropy decoding the compressed image to reconstruct the watermarked quantized DCT coefficients;
(e) restoring the quantized DCT coefficients by extracting an authentication code from the restored watermarked quantized DCT coefficients;
(f) reconstructing the original image through inverse quantization, inverse DCT, and color space change of the reconstructed quantized DCT coefficients;
(g) regenerating an authentication code from the restored original image; And
(h) comparing the authentication code extracted in step (e) with the authentication code regenerated in step (g) to confirm the integrity of the image; and the integrity of the image using a reversible watermarking technique. Authentication method. According to claim 1, wherein the authentication code generation of the step (b),
Obtaining the quantized DCT coefficients through the step (a), and selecting the n highest values in zig-zag order from the obtained quantized DCT coefficients to generate the image integrity authentication method using a reversible watermarking technique. According to claim 1, wherein the authentication code insertion of step (b),
Shifting positive values of the quantized DCT coefficients and inserting an authentication code between 0 and the shifted positive value. The method of claim 1, wherein the authentication code regeneration of the step (g),
DCT transform and quantize the original image reconstructed in step (f) to extract and generate quantized DCT coefficients. The process of claim 1, wherein the series of steps is
An image integrity authentication method using a reversible watermarking technique, characterized in that the process of converting the original image of the upper block to the lower block unit image and then compressing and restoring the original image.

Images