KR101220147B1 - Differential histogram-based reversible watermarking method through row-column cross scanning - Google Patents

Abstract

In order to increase the insertion capacity through repetitive insertion of a message-inserted image, the difference between matrix cross-scanning methods can be used to maximize the efficiency of locality and repetitive insertion when repetitive insertion is performed by changing the scan direction for calculating the histogram. A histogram based reversible water marking method is provided. The difference image histogram is gradually constructed by scanning rows or columns first according to the round number of times of repeated insertion for the original image. Shift histogram bins that are not used for insertion to make room for insertion. The difference histogram is modified according to the bits of the message to be inserted to insert the message. A hidden image is generated by reflecting the modified difference histogram in a row or column priority direction according to the round insertion number round value. The difference image histogram is gradually formed by scanning the hidden image by row or column priority according to the round insertion number round value. Scan the message insertion space to detect the inserted message. Restore the modified difference histogram to insert the message. Restore histogram bins that have been shifted for space. The original image is reconstructed in a row or column priority direction according to the round insertion number round value using the reconstructed difference histogram.

Description

Differential histogram-based reversible watermarking method through row-column cross scanning

The present invention relates to a reversible watermarking method, and more particularly, to a difference value histogram-based reversible watermarking method.

Data hiding technology as a post-security technology for content is a technology that invisibly inserts confidential information into digital content such as music, video, video, electronic documents, and animations, and provides proof of ownership, copyright protection, broadcasting monitoring, and content authentication. It is used for various purposes. Encryption technology is also a method for protecting digital content, but it only guarantees protection in the content distribution process, and once decrypted content can no longer be protected, it is not enough to prove the integrity of the content. On the other hand, data hiding technology is a good way to protect the content after it is distributed because it can satisfy the requirements of various insertion capacity, perceptual transparency, robustness, confidentiality, and computational complexity depending on the application.

Digital watermarking, a representative data concealment technology, can invisibly insert information such as metadata about original contents, authentication codes for integrity verification, copyright information, and the like into the contents. In order to conceal data in content, modification of the original content is inevitably inevitable. In application fields such as medical imaging, military imaging, legal evidence, telemetry, and works of art, an original image without any damage is required. This is because the degree of change is minimal and human perceptions may not be recognized at all, which can affect the right decision and become a legal issue. Reversible watermarking can be used to remove the message from the watermarked content and then restore it completely to the original content, which can be used as an excellent way to verify the integrity of the content, to prove tampering, and to protect copyrights. This reversible watermarking technique is a message concealment means that maintains perceptual transparency to digital content and restores it to its original state without damaging watermarks. Can be used. The reversible watermarking algorithm should be perceptually transparent so that the modified image is not significantly different from the original image by inserting a message, while providing sufficient insertion capacity to meet the needs of the application. In addition, the computational complexity should not be high so that it can be applied to small devices with relatively low resources and computational power, and reversibility must be ensured to allow full restoration to the original after extracting a message.

Recent reversible watermarking studies have sought to maximize insertion capacity while minimizing quality degradation. Existing histogram-based studies can increase the insertion capacity by iteratively inserting (reinserting) messages with embedded images. However, once the message is inserted, the height of the maximum point in the histogram is reduced by about half, so the efficiency of repetition is drastically lowered as the level of repetition increases.

Reversible watermarking studies insert messages using different image features to provide perceptual transparency and full reversibility. Fridrich et al. And Celik et al. Use a lossless compression method to compress the bit plane and insert a message into the empty space. Yang et al. And Xuan et al., Lee et al. Inserted in the count. In addition, Tian, Alattar, Kamstra et al. And Thodi et al. Used the difference value expansion method that extends and inserts the characteristic values of the original image. Recently, researches on histogram-based methods that can obtain high insertion capacity, which are not high in computational complexity, are being actively conducted. According to algorithms, histograms for image brightness values or histograms for difference values are used.

Reversible watermarking algorithms using the brightness value histogram correction technique insert a piece of data by shifting the bin around the maximum point into which the data is to be inserted, freeing space, and shifting the maximum point to the left or right according to the bits of the message to be inserted. . Therefore, in order to obtain high data insertion capacity, it is necessary to secure many maximum points. However, the more the maximum points are used, the more severe the distortion of the original image, the amount of overhead information and the complexity of the algorithm increase. On the other hand, since the difference histogram can be used to obtain a higher insertion capacity than the use of the brightness histogram even if only one maximum point is used, recent reversible watermarking studies tend to use the difference histogram.

Difference Histogram

Algorithms using difference histograms are calculated using the difference in brightness from adjacent pixels. As shown in FIG. 1, the maximum value of the brightness histogram is 2,751 in the Lena image, but the maximum value of the difference histogram is 22,252. Having a high maximum means that a higher insertion capacity can be obtained with less distortion, so it is more advantageous to use a difference histogram. In addition, the method using the brightness value histogram causes additional overhead for storing the original maximum point position since the position of the maximum point changes after data insertion. However, in the method using the difference histogram, since the difference values are concentrated around 0 due to the locality characteristic of the image data that the change between adjacent pixels is small, the position of the maximum point can be fixed and processed, so the overhead of the insertion position information Has the advantage that it is not necessary.

When the watermark is inserted using the histogram shifting method, the difference value histogram is used instead of the brightness value histogram in order to secure a higher maximum point as shown in FIG. 1. The higher the maximum point, the higher the insertion capacity. The difference histogram uses a local feature that the difference between adjacent pixels is not significant, so that the maximum point is much higher than the brightness histogram. Reversible watermarking algorithms using difference histograms show various results depending on how the difference histogram is constructed.

Lee et al. Constructed a histogram using the difference between odd and even rows of an image. Therefore, the actual difference image used for message insertion is 50% larger than the original. Since the area modified for message insertion is less than 50%, the quality of the hidden image is high, but the insertion capacity is not large. Lin et al. Divided the image into blocks of size A x B and calculated the difference between the current pixel and the next pixel in each block to generate the difference image of size A x (B-1). The size of the total difference image used to insert the message depends on the block size. In the case of a 4 x 4 block, the size of the difference image is 75% larger than the original. Tsai et al. Divided the image into blocks of size N x N and constructed a histogram using the difference between the reference pixel and the rest of the pixels in each block. However, since the difference between adjacent pixels is not used, the maximum value of the histogram is not concentrated around zero. Therefore, like the method using the brightness value histogram, there is an overhead of separately transmitting positional information about the maximum and zero points. Kim et al. Subsampled the images at regular intervals and constructed a histogram using the difference between the reference subimage and the remaining subimages. This method does not modify the reference sub-images and the remaining unsub-sampled regions. Therefore, when the subsampling interval is 3, the difference image used for message insertion is about 88% larger than the original image.

Yeo et al. Constructed a histogram by gradually calculating the difference between the current pixel and the immediately preceding pixel for the entire image. This method shows the highest insertion capacity compared with previous studies because the difference image size used for message insertion is the same size as the original.

Watermarking algorithms must provide sufficient insertion capacity to meet the needs of the application. If the size of the message to be inserted is larger than the actual insertable capacity, histogram-based studies have determined that the insertion cannot be made or the method of increasing the insertion capacity by inserting the message repeatedly is selected. have. However, once the message is inserted, the height of the maximum point in the histogram is reduced by about half, so the efficiency of repetitive insertion decreases rapidly as the level of repetition increases. Hereinafter, the maximum value decrease due to the repetitive insertion will be described.

In the conventional technique, the highest peak among the methods used to calculate the difference is Yeo et al. Gradual difference histogram method is to use the difference between the previous pixel and the current pixel. In this method, since the image is scanned one pixel at a time, the difference value is calculated in the sequential scanning direction as shown in FIG. 2.

It is necessary to reconstruct the difference value histogram in order to repeat the insertion in order to increase the insertion capacity of the image into which the message is inserted by the sequential scanning method. To this end, when the difference histogram is calculated by the sequential scanning method, the height of the maximum point in the histogram is reduced to about half as shown in FIG. 3. Therefore, as the level of repetition increases, a problem arises in that the efficiency drops rapidly.

The present invention has been made to improve the problems of the prior art as described above, and when repetitive insertion is performed by changing the scan direction for calculating the histogram when the insertion capacity is increased by repetitive insertion for a message inserted image. The purpose of the present invention is to provide a reversible watermarking method based on difference between histograms and matrix cross-scanning methods that can maximize the efficiency of locality and repeated insertion.

The matrix cross-scanning difference histogram-based reversible watermarking method according to the present invention comprises the steps of: (i) gradually constructing a difference value histogram by scanning rows or columns first with respect to a round number of repetition insertion times for an original image; (ii) shifting histogram bins that are not used for insertion to free up insertion space; (iii) inserting the message by modifying the difference histogram according to the bits of the message to be inserted; (iv) generating a hidden image reflecting the corrected difference histogram in a row or column priority direction according to the round insertion number round value; (v) gradually constructing a difference value histogram by scanning rows or columns first with respect to the hidden insertion number round value of the hidden image; (vi) scanning the message insertion space to detect the inserted message; (vii) restoring the modified histogram that has been modified to insert the message; (viii) restoring histogram bins that have been shifted for space; And (ix) restoring the original image in a row or column priority direction according to the repeated insertion number round value using the restored difference histogram.

As described above, according to an experimental result of comparative analysis of various images, the proposed algorithm can obtain a high insertion capacity through efficient repetitive insertion while maintaining low distortion with complete reversibility. In order to maximize the locality of the modified image once inserted, the performance is improved by applying a matrix cross-scan technique that adjusts the scan direction to calculate the difference histogram each time the repetition step is performed. Compared to the conventional method using only row-first scanning, the proposed algorithm improves not only the insertion capacity but also the quality performance.

1 is a graph illustrating a histogram of brightness values and a difference histogram of a Lena image.
2 is a diagram illustrating sequential scanning.
3 are graphs illustrating changes in a difference value histogram after message insertion on a Lena image.
4 is a diagram illustrating a histogram bin used to insert a message according to an embodiment of the present invention.
5 is a diagram illustrating a message insertion process in a matrix cross-scan method difference value histogram-based reversible water marking method according to an embodiment of the present invention.
6 is a diagram illustrating row priority sequential scanning and column priority sequential scanning according to an embodiment of the present invention.
7 are graphs illustrating a difference value histogram comparison result of row-first and column-first sequential scanning after message insertion according to an embodiment of the present invention.
8 is a diagram illustrating matrix intersection scanning according to an increase in the number of repetitions according to an embodiment of the present invention.
9 is a flowchart illustrating an iterative insertion procedure using matrix cross scanning according to an embodiment of the present invention.
10 is a diagram illustrating a histogram modification procedure in a message detection and restoration process according to an embodiment of the present invention.
11 shows an 8-bit grayscale 512 × 512 experiment image.
12 are graphs illustrating a change in scan direction of a Lena image when 2/4/6/8 repetition is inserted.
FIG. 13 is a graph illustrating scanning direction changing performance of a Lena image when 0/3/6/9 repetitive insertion is performed.
FIG. 14 is a graph illustrating scanning direction changing performance (average value) for the entire image when 2/4/6/8 repetition is inserted.
FIG. 15 is a graph illustrating scanning direction changing performance (average value) of an entire image when 0/3/6/9 repeated insertion is performed.
FIG. 16 is a graph illustrating a change in performance with respect to an insertion level and an increase in the number of iterations of matrix intersection scanning according to an embodiment of the present invention.

Hereinafter, a method of histogram-based reversible water marking based on a matrix cross scanning scheme difference value according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear. The following terms are defined in consideration of the functions of the present invention, and may be changed according to the intentions or customs of the user, the operator, and the like. Therefore, the definition should be based on the contents throughout this specification.

The present invention proposes a method for maximizing the efficiency of locality and repetition by adjusting the scan direction for calculating the difference value. The message insertion and detection algorithm uses a reversible watermarking algorithm using progressive difference histogram shifting, which has been studied by the inventor of the present invention.

Message insertion process

The capacity of the message to be inserted and the quality of the watermark image are adjusted to the insertion level L according to the requirements of the application field. According to the value of L starting from 0, the bin used for message insertion in the difference histogram is shown in FIG. 4 from {(-L-1) to (+ L)} around the 0th bean.

When the insertion level L is 1, a procedure of modifying the histogram in the process of inserting a message is illustrated in FIG. 5.

Efficient iterative insertion with matrix cross scanning

The reason why the difference histogram is reduced by about half after inserting the message is because it inserts the message by unfolding the histogram bin around the maximum point. In order to create an image with a message inserted, the histogram must be shifted to change the pixel value by the corrected difference value, thereby increasing the difference value, thereby degrading locality. However, when scanning sequentially for re-insertion, scanning in column-major order rather than row-major order can significantly compensate for the locality deterioration of sequential scanning. This is because column-first sequential scanning shows higher locality than row-first sequential scanning since the pixel value in the horizontal direction is increased and decreased in the vertical direction as well.

First, when a user inputs a round insertion number round value and an original image, the R value is increased by 1 from 1 to the round value, and the next insertion process is repeated.

6 illustrates the concept of row first and column first sequential scanning. In FIG. 7, the difference values histograms of row-first sequential scanning of originals are compared with the difference values histograms of row-first and column-first sequential scanning after inserting a message. As can be seen from FIG. 7, when row-first sequential scanning is repeatedly applied to an image in which a message is inserted through row-first sequential scanning, the average height is 47% higher than that of the original. You can get height. Therefore, it can be seen that the use of alternating row-first and column-first sequential scanning every time iterative insertion can minimize local deterioration.

In order to change the direction of the sequential scanning according to the repetitive insertion, the row-first sequential scanning and the column-first sequential scanning are alternately used according to the number of repetition inserts R, and the difference histogram is calculated. In the present invention, this is defined as row-column cross scanning. The principle of matrix cross scanning is shown in FIG. 8, and FIG. 9 shows an iterative insertion procedure.

If R is odd, the original image is scanned row-first and even-numbered column first to construct a difference histogram gradually (Fig. 5 (a)). The bins not used for insertion are shifted to secure the insertion space (Fig. 5 (b)).

Next, the difference value histogram is modified according to the bits of the message to be inserted to insert the message (FIGS. 5 (c) and 5 (d)). If R is odd, a hidden image reflecting the modified histogram is generated in a row-first direction and in an even-numbered column.

Message Detection and Recovery Algorithm

When the insertion level L is 1, a procedure of correcting the histogram in the process of detecting the message and restoring the original image is illustrated in FIG. 10. First, when a user inputs a round insertion number and a hidden image, the R value is decreased by 1 from the round value to 1 and the next restoration process is repeated.

If R is odd, the hidden image is scanned with row priority and even with column priority to construct a difference value histogram gradually (Fig. 10 (a)). The inserted message is detected by scanning the message insertion space (Fig. 10 (b)).

Next, the difference value histogram which has been modified to insert the message is restored (Fig. 10 (c)). The bins that have been shifted to secure the space are restored (FIG. 10 (d)). Finally, using the reconstructed difference histogram, the original image is reconstructed in the row-first direction when R is odd and in the column-first direction by even numbers (FIG. 10 (e)).

Experiment and performance evaluation

The images used in the experiments according to the invention are seven 8-Bits grayscale 512 x 512 images from the University of Southern California-Signal & Image Processing Institute (USC-SIPI) image database and are shown in FIG. 11. The effective insertion capacity is a capacity obtained by subtracting overhead information, and the image quality is measured by PSNR (dB) calculated through Equations 1 and 2 below.

In Equation 1, M and N are horizontal and vertical sizes of an image, respectively, p _{(i, j)} is a pixel value of an original image, p ' _{(i, j)} is a pixel value of an image to which a mark is inserted. 2 to n is the number of bits needed to represent a pixel.

Fig. 12 to Fig. 13 show performance comparisons when only the row first sequential scanning method is applied to the Lena image and when the matrix cross scanning method proposed in the present invention is applied. 12 is a result of fixing the number of repetition insertions R to 2/4/6/8 and changing the insertion level from 0 to 9, and FIG. 13 to fix the insertion level to 0/3/6/9 and repeat. It shows the result when the frequency is changed from 1 to 10. Since the results for all the experimental images show a pattern very similar to the Lena images shown in FIGS. 12 and 13, the present invention does not include the results for the remaining experimental images. Average values for the results of the dog images are shown in FIGS. 14 and 15. Each point in the graph represents the results of experiments at a specific insertion level and the number of repeated insertions. The results of experiments in which the overhead information is larger than the insertable capacity are not shown in the graph. Experimental results show that it is more efficient to insert repeatedly using the matrix cross-scanning method proposed in the present invention than to use only one scanning method as in the conventional techniques. It can be seen that it is quite relaxed.

Compared with the conventional row-first scanning method, the performance improvement of the quality (PSNR) and insertion capacity (bpp) scales through the matrix cross-scanning method proposed by the present invention is performed twice for the insertion level L = {0,2,4}. The results obtained by inserting four times are shown in Tables 1 and 2. It can be seen that the higher the stage of repetition, the greater the insertion dose, and the higher the quality, despite the more insertions.

Insertion condition Lena Baboon Boat Airplane Aerial Tank Trucks Average 0L 2R 37% 24% 29% 20% 16% 12% 13% 22% 2L 2R 23% 17% 19% 12%  9% 11% 12% 15% 4L 2R 17% 13% 16%  8%  5%  8% 10% 11% 0L 4R 63% 50% 51% 39% 36% 36% 38% 44% 2L 4R 40% 38% 35% 25% 24% 29% 32% 31% 4L 4R 30% 30% 30% 21% 19% 22% 27% 25%

Insertion condition Lena Baboon Boat Airplane Aerial Tank Trucks Average 0L 2R 4% 5% 4% 4% 5% 5% 5% 5% 2L 2R 5% 6% 5% 5%  6% 5% 6% 5% 4L 2R 6% 6% 6%  6%  7%  5% 6% 6% 0L 4R 5% 6% 5% 5% 6% 5% 6% 5% 2L 4R 6% 7% 6% 6% 7% 6% 7% 6% 4L 4R 7% 8% 7% 7% 8% 6% 7% 7%

FIG. 16 compares the performance of repeated insertion by increasing the insertion level and repeating insertion by increasing the number of repetitions when repeated insertion using matrix intersection scanning. 16 (a) and 16 (b) show results for the Lena image, and FIGS. 16 (c) and 16 (d) show average values for the results of seven images. As can be seen from the results, a higher insertion level and repeated insertions with a smaller number of times perform slightly better on the insertion capacity and quality scale than lower insertion levels and a larger number of repeated insertions. Therefore, it can be seen that it is excellent not only in the insertion capacity but also in the quality in that it is repeatedly inserted several times by selecting a method capable of inserting a lot of capacity in one insertion process.

Although the present invention has been described as a specific preferred embodiment, the present invention is not limited to the above-described embodiments, and the present invention is not limited to the above-described embodiments without departing from the gist of the present invention as claimed in the claims. Anyone with a variety of variations will be possible.

Claims (5)

(i) progressively constructing a difference value histogram by scanning rows or columns first with respect to the round insertion number round value of the original image;
(ii) shifting histogram bins that are not used for insertion to free up insertion space;
(iii) inserting the message by modifying the difference histogram according to the bits of the message to be inserted;
(iv) generating a hidden image reflecting the corrected difference histogram in a row or column priority direction according to the round insertion number round value;
(v) gradually constructing a difference value histogram by scanning rows or columns first with respect to the hidden insertion number round value of the hidden image;
(vi) scanning the message insertion space to detect the inserted message;
(vii) restoring the modified histogram that has been modified to insert the message;
(viii) restoring histogram bins that have been shifted for space; And
and (x) restoring the original image in a row or column priority direction according to the repeated insertion number round value using the restored difference histogram. The method according to claim 1, wherein, in step (i), a row-first scanning is performed first if the number of rounds of repeated insertion is odd for the original image, and column-first scanning if the even number is even. The method according to claim 1, wherein in step (iv), if the number of rounds of the repetition insertion is odd is row-wise, and if is even, column-cross scanning method difference value histogram-based reversible water for generating a hidden image reflecting the modified histogram Marking method. The method according to claim 1, wherein in step (v), the row-first scanning is performed first, if the number of rounds of the repetition insertion number is odd, and column-first, if the even number is even. 2. The method of claim 1, wherein, in step (ix), the original image is restored in a row-first direction when the number of rounds of repeated insertion is odd and in a column-first direction when an even number is even. 3.

Images