TWI447669B - System and method for removing watermarks from an image - Google Patents

Description

Watermark information removal system and method

The present invention relates to a digital image processing system and method, and more particularly to a system and method for removing watermark information in a digital image.

In the information age of the Internet's booming information, the speed of digital information transmission and reproduction is quite fast. In order to prevent the intellectual property rights of digital media from being infringed, especially the commercial image and music, it is one of the problems that need to be addressed at present. . Therefore, many techniques for watermark information hiding are widely proposed to protect or verify the owner's copyright information. With the rise of e-commerce, business practices on the Internet have become more and more frequent, and the increasing number of Internet infringements has made digital media's intellectual property protection issues more and more important. For example, many image processing software have also added watermarks. Features to protect the user's efforts. However, since there are techniques for protection, there are of course ways to try to remove the watermark information.

At present, the watermarking attack removes the watermark information hidden in the image, and the image after the watermark information is removed can retain the commercial value of the original media, so as to successfully achieve the purpose of removing the watermark information. However, although the current watermark removal technology can remove the hidden watermark information from the original image, the quality of the image after the watermark information is removed cannot be guaranteed, so that the image distortion after the watermark information is removed cannot be Preserve the commercial value of the original video media.

In view of the above, it is necessary to provide a watermark information removal system and method, which can not only remove the watermark information hidden in the watermark image, but also keep the image from being distorted.

A watermark information removal system is installed and operated in an electronic device, the electronic device including a memory. The system includes: an image access module for reading a watermark image from the memory; and an image decomposition module for decomposing the watermark image into a corresponding number of feature graphics according to the size of the watermark image And each feature graphic is sequentially numbered; the feature graphic identification module is configured to identify the range of the graphic number to be removed from the decomposed feature graphic; the feature graphic screening module is used to calculate the identified The feature values of all feature points in each feature graphic in the range of the graphic number, the feature values of all the feature points are added to obtain the total feature value of the feature graphic, and the watermark information is filtered according to the total value of the feature image. The feature graphic; the watermark information removal module is configured to delete the filtered feature graphic and combine the remaining feature graphics into a combined image.

A method for removing watermark information is applied to an electronic device, and the electronic device includes a memory. The method comprises the following steps: (a) reading a watermark image in the memory; (b) decomposing the digital watermark image into a corresponding number of feature graphics according to the size of the watermark image, and each The feature graphic is sequentially numbered in the figure; (c) the feature image of the feature graphic to be removed is identified in the decomposed feature graphic; (d) the watermark information is filtered out within the range of the identified graphic number Feature graphics; (e) deleting the filtered feature graphics and combining the remaining feature graphics into a combined image; (f) checking whether the quality of the combined image is distorted; and (g) storing the combined image without distortion in the memory in.

Compared with the prior art, the watermarking information removal system and method can hide the watermark hidden in the watermark image by a singular value decomposition algorithm and a differential distortion algorithm based on the feature image. The information is removed, in addition to the watermark information hidden in the watermark image, the original image quality can be preserved.

1‧‧‧Watermark Information Removal System

11‧‧‧Image Access Module

12‧‧‧Image Decomposition Module

13‧‧‧Characteristic Graphic Identification Module

14‧‧‧Characteristic graphic screening module

15‧‧‧Watermark Information Removal Module

16‧‧‧Image quality inspection module

2‧‧‧ Processor

3‧‧‧ memory

S11‧‧‧Read watermark image from memory

S12‧‧‧ Decompose watermark images into feature graphics using SVD algorithm

S13‧‧‧Use JND algorithm to identify the range of graphic numbers

S14‧‧‧ Screened out feature graphics with watermark information

S15‧‧‧Delete the selected feature graphics and combine a combined image

S16‧‧‧ Calculate the PSNR value of the combined image

S17‧‧‧Determination of whether the combined image is distorted based on the PSNR value

S18‧‧‧Storing combined images into memory

1 is an application environment diagram of a preferred embodiment of the watermarking information removal system of the present invention.

2 is a flow chart of a preferred embodiment of the method for removing watermark information according to the present invention.

FIG. 3 is a detailed flowchart of step S13 in FIG. 2.

FIG. 4 is a detailed flowchart of step S14 in FIG. 2.

Referring to FIG. 1, an application environment diagram of a preferred embodiment of the watermarking information removal system 1 of the present invention is shown. The watermark information removal system 1 is installed and operates in an electronic device including a processor 2 and a memory 3. In this embodiment, the electronic device is a mobile phone. In other embodiments, the electronic device may be a computer, a PDA, or other electronic device having image processing functions. The processor 2 is configured to run the watermark information removal system 1 to remove the watermark information hidden in the watermark image from the watermark image. The memory 3 is used for storing a watermark image with watermark information and a combined image obtained by removing the watermark information.

The watermark information removal system 1 is configured to read a watermark image from the memory 3, decompose the watermark image into a certain number of feature graphics (Eigen-image), and find and remove the watermark information. The feature graphic combines the remaining feature graphics into a combined image after the watermark information is removed, and checks the quality of the combined image, and stores the combined image in the memory 3. In this embodiment, the The watermark information removal system 1 includes an image access module 11 , an image decomposition module 12 , a feature graphics authentication module 13 , a feature graphics screening module 14 , a watermark information removal module 15 , and an image quality inspection module 16 . .

The image access module 11 is configured to read a watermark image from the memory 3, and store the combined image obtained by removing the watermark information from the watermark image into the memory 3. The watermark image is hidden with corresponding watermark information. For example, the watermark image hides the copyright information of the protection or verification owner.

The image decomposition module 12 is configured to decompose the watermark image into a certain number of feature graphics according to the size of the watermark image by using a Singular Value Decomposition (SVD) algorithm, and each feature graphic is The graphic number is in order. The image decomposition module 12 can decompose the watermark image into 64 feature graphics, 128 feature graphics, or a greater number of feature graphics. In this embodiment, the original image size is 128*128 primitives, and the image decomposition module 12 decomposes the watermark image into 128 feature graphics, and each of the feature graphics has a corresponding graphic number, for example, the feature graphic once. It is numbered as the first, second, third...128 feature graphics. The SVD algorithm is an algorithm for extracting watermark information by calculating a maximum singular value point in a watermark image by a quantization strategy. Due to the feature pattern decomposed by the SVD, the closer the feature image is to the original image, the closer it is to the original image. Therefore, if 128 feature graphics are recombined, the combined images will be the same as the original image. If the first 126 feature graphics are combined, the combined image and original image differences are indistinguishable to the user's naked eye.

The feature pattern discriminating module 13 is configured to identify a range of pattern numbers to be removed by using a Just Noticeable Distortion (JND) algorithm on the decomposed feature pattern. The JND algorithm is a calculation feature map An algorithm for determining whether the feature graphic contains watermark information in the feature value corresponding to the feature point in the shape.

The feature graphic screening module 14 is configured to calculate feature values of all feature points in each feature graphic within the identified graphic number range, add the feature values of all the feature points to obtain the feature total value of the feature graphic, and The size of the total value of the feature filters out the feature graphic with the watermark information. If the total value of the feature of a certain feature graphic is larger, the more watermark information is hidden by the feature graphic; otherwise, if the total value of the feature of a certain feature graphic is smaller, the feature graphic is hidden. The less the watermark information.

The watermark information removal module 15 is configured to delete the filtered feature graphics with watermark information and combine the remaining feature graphics into one image.

The image quality check module 16 is configured to calculate a peak signal to noise ratio (PSNR value) of the combined image, and determine whether the quality of the combined image is based on whether the PSNR value exceeds a preset PSNR allowable value. distortion. In this embodiment, the image quality inspection module 16 obtains the PSNR value of the combined image by calculating the ratio of the primitive value of the watermark image to the primitive value of the combined image. Generally, the user can set the PSNR allowable value of the watermark image according to the size of the primitive value of the watermark image. For example, the user can set the PSNR allowable value of the watermark image to 0.8. That is to say, if the PSNR value of the combined image is greater than or equal to 0.8, the distortion of the combined image relative to the watermark image is small, and the difference from the original image is indistinguishable to the user's naked eye.

Referring to FIG. 2, it is a flowchart of a preferred embodiment of the method for removing watermark information of the present invention. In this embodiment, the watermarking information removal method removes the watermark information hidden in the watermark image by an SVD decomposition technique and a JND technology based on the feature image, except that the image can be hidden. Watermark information in watermark images In addition to the removal, the original image quality can be preserved.

The image access module 11 reads a watermark image in the memory 3 (step S11). The watermark image is hidden with corresponding watermark information. For example, the watermark image hides the copyright information of the protection or verification owner.

The image decomposition module 12 uses the SVD algorithm to decompose the watermark image into a certain number of feature graphics according to the primitive size of the watermark image, and sequentially numbers each feature graphic (step S12). In this embodiment, the size of the watermark image is 128*128, and the image decomposition module 12 decomposes the watermark image into 128 feature graphics, and each of the feature graphics has a corresponding graphic number. For example, the feature graphic is numbered at a time. It is the characteristic figure of No. 1, 2, 3...128.

The feature pattern discriminating module 13 uses the JND algorithm to identify the pattern number range of the feature pattern to be removed from the decomposed feature pattern (step S13). In this embodiment, the feature pattern discriminating module 13 identifies the number of the pattern numbers to be removed from the 128 feature patterns from the 50th to the 100th sheets. How the feature graphic screening module 13 utilizes the JND algorithm to identify the feature number that needs to be removed for feature pattern recognition will be described in detail in FIG.

The feature graphics screening module 14 filters out the feature graphics having the watermark information within the identified number of graphics numbers (step S14). How the feature graphic screening module 14 screens out the feature graphic with the watermark information will be described in detail in FIG.

The watermark information removing module 15 deletes the filtered feature graphic with the watermark information, and combines the remaining feature graphics into one combined image after the attack (step S15).

The image quality checking module 16 calculates a primitive value of the watermark image and a map of the combined image The ratio of the values gives the PSNR value of the combined image (step S16). The image quality checking module 16 determines whether the quality of the combined image is distorted by determining whether the PSNR value is less than a PSNR allowable value set by the user (step S17). Generally, the user can set the PSNR allowable value of the watermark image according to the size of the primitive value of the watermark image. For example, the user can set the PSNR allowable value of the watermark image to 0.8. That is to say, if the PSNR value of the combined image is greater than or equal to 0.8, the distortion of the combined image relative to the watermark image is small, and the difference from the original image is indistinguishable to the user's naked eye.

In step S17, if the quality of the combined image is distorted, that is, the PSNR value of the combined image is less than the PSNR allowable value set by the user, the flow returns to step S13 to re-authenticate the range of the graphic number to be removed to hide the watermark. The watermark information in the image is removed. If the quality of the combined image is not distorted, that is, the PSNR value of the combined image is greater than or equal to the PSNR allowable value set by the user, the image access module 11 stores the combined image without distortion in the memory 3 (step S18). Generally, if the combined image has a small degree of distortion with respect to the watermark image, it is considered that the difference between the combined image and the original image is indistinguishable to the user's naked eye and can be accepted by the user.

Referring to FIG. 3, a detailed flowchart of step S13 in FIG. 2 is shown. The refinement flow method illustrates how to use the JND algorithm to identify the feature number range of the feature graphics that need to be removed from the feature graphics.

The feature pattern discriminating module 13 converts the watermark image into a JND image using the JND algorithm (step S131). The feature pattern discriminating module 13 calculates a JND value of each primitive point in the JND image, which is expressed as "J _i (x, y)" (step S132), and calculates all the primitive points in the JND image. The average JND value, which is expressed as "J _avg " (step S133).

The feature pattern discriminating module 13 compares the JND value J _i (x, y) and the average JND value J _avg of each of the primitive points in the feature pattern (step S134). The feature pattern discriminating module 13 selects a feature point of the feature point in which the JND value J _i (x, y) of the primitive point is larger than the average JND value J _avg as the feature point of the feature pattern (step S135).

The feature pattern discriminating module 13 compares the JND values corresponding to the arbitrary feature points in each of the feature patterns (step S136). The feature pattern discriminating module 13 selects the feature pattern whose feature point is close to the corresponding JND value in the feature pattern, and places the feature pattern into the figure number range of the feature pattern to be removed (step S137). In this embodiment, if a certain feature point FP (10, 18) has a large difference in feature values in the 0th to 49th feature patterns, the difference in feature values in the 50th to 100th feature patterns is relatively close. However, the feature value difference in the 101st to 128th feature graphics is also larger than the pair, so the feature graphic discriminating module 13 puts the graphic number range into the 50th to 100th feature graphic into the graphic of the feature graphic to be removed. Within the number range.

Referring to FIG. 4, a detailed flowchart of step S14 in FIG. 2 is shown. The refinement process method illustrates how to use the JND algorithm to filter out feature patterns with watermark information. The feature pattern marked below within the range of the graphic number is the feature graphic "E _i ".

The feature graphic filtering module 14 calculates the primitive value of each primitive point in the feature graphic E _i , which is expressed as "E _i (x, y)" (step S141). At the same time, the feature graphic screening module 14 calculates the average primitive value of all the feature points in the feature graphic E _i , which is expressed as "Ei_mean" (step S142).

The feature graphic screening module 14 determines whether the primitive value E _i (x, y) of each primitive point in the feature graphic E _i is greater than the average primitive value E _{i_mean} (step S143). If the primitive value E _i (x, y) of the primitive point is less than or equal to the average primitive value E _{i_mean} , the feature graphic screening module 14 converts the primitive value E _i (x, y) in the feature graphic E _i It is a binary value "0" (step S144). If the primitive value E _i (x, y) of the primitive point is greater than the average primitive value E _{i_mean} , the feature graphic screening module 14 converts the primitive value E _i (x, y) in the feature graphic E _i into The binary value is "1" (step S145). The feature graphic screening module 14 records the number of feature points whose value is "1" in each feature pattern E _i (step S146), and accumulates the number of feature points to obtain the total feature value of each feature pattern E _i . It is expressed as "BC _i " (step S147).

The feature graphic screening module 14 sorts each feature total value BC _i in order of largest to small (step S148). Finally, the feature graphic screening module 14 selects the feature graphic (labeled as the feature graphic “E _r ”) corresponding to the largest feature total value BC _i , and the feature graphic E _r is the feature graphic E _{i that} needs to be removed. The figure number is removed within the range (step S149). Generally, if the total value of the feature point feature characteristic pattern BC _i E _r is larger, the characteristic pattern E _r of the more hidden watermark information.

The present invention has been described above in terms of preferred embodiments, and is not intended to limit the invention. The scope of the present invention is defined by the scope of the appended claims, unless otherwise claimed.

S11‧‧‧Read watermark image from memory

S12‧‧‧ Decompose watermark images into feature graphics using SVD algorithm

S13‧‧‧Use JND algorithm to identify the range of graphic numbers

S14‧‧‧ Screened out feature graphics with watermark information

S15‧‧‧Delete the selected feature graphics and combine a combined image

S16‧‧‧ Calculate the PSNR value of the combined image

S17‧‧‧Determination of whether the combined image is distorted based on the PSNR value

S18‧‧‧Storing combined images into memory

Claims (9)

A watermark information removal system installed and operated in an electronic device, the electronic device comprising a memory, the system comprising: an image access module for reading a watermark image from the memory; the image decomposition module The group is configured to decompose the watermark image into a corresponding number of feature graphics according to the size of the watermark image, and each of the feature graphics is sequentially numbered; the feature graphic identification module is used to decompose the feature The graphic identification needs to remove the range of the graphic number, including the steps of: converting the watermark image into a differential distortion image by using an identification distortion algorithm; calculating the distortion value of each primitive point in the identified distortion image; Identifying the average distortion value of all the primitive points in the distorted image; comparing the distortion value and the average distortion value of each primitive point; and using the primitive point whose distortion value is greater than the average distortion value as the feature point of the feature graphic; Corresponding to the distortion value corresponding to each feature pattern in the feature pattern; and selecting the feature pattern of the feature point corresponding to the distortion value in the feature pattern, and The feature graphic is placed in the range of the graphic number of the feature graphic to be removed; the feature graphic filtering module is configured to calculate the feature value of all the feature points in each feature graphic within the identified graphic number range, and add up The feature values of all the feature points obtain the feature total value of the feature graphic, and the feature graphic with the watermark information is filtered according to the total value of the feature; and the watermark information removal module is used to delete the selected feature The graphics combine the remaining feature graphics into a single combined image. The watermarking information removal system of claim 1, wherein the system further comprises an image quality checking module, configured to calculate a peak noise ratio of the combined image, and determine the combined image according to the peak noise ratio Whether the quality is distorted. The watermarking information removal system of claim 2, wherein the image access module is further configured to store the combined image without distortion into the memory. The watermarking information removal system of claim 1, wherein the image decomposition module decomposes the watermark image into a feature graphic by using a singular value decomposition algorithm, and the singular value decomposition algorithm is a The algorithm for extracting the watermark information hidden in the watermark image by calculating the maximum singular value point in the watermark image by the quantization strategy. The watermarking information removal system of claim 1, wherein the feature pattern discriminating module uses a discriminant distortion algorithm to identify a range of pattern numbers to be removed, and the discriminant distortion algorithm is a The algorithm for determining whether the feature graphic contains the watermark information is determined by calculating the size of the feature value corresponding to the feature point in the feature graphic. A watermark information removal method is applied to an electronic device, the electronic device comprising a memory, the method comprising the steps of: (a) reading a watermark image in the memory; (b) according to the watermark image The size of the primitive is decomposed into a corresponding number of feature graphics, and each feature graphic is sequentially numbered in sequence; (c) the feature image is decomposed to identify the graphic number range of the feature graphic to be removed. Wherein the step (c) comprises: converting the watermark image into a discriminant distortion image by using a discriminant distortion algorithm; calculating a distortion value of each primitive point in the discriminative distortion image; and calculating the discriminant distortion image The average distortion value of all the primitive points; the distortion value and the average distortion value of each primitive point are compared; the primitive point whose distortion value is larger than the average distortion value is used as the feature point of the feature graphic; the comparison feature point is a corresponding distortion value in each feature graphic; and selecting a feature graphic whose feature point is close to a corresponding distortion value in the feature graphic, and placing the feature graphic into a special feature to be removed Graphic pattern numbers within the range: (d) have been identified in the screening of the pattern number range having a characteristic pattern of the watermark information; (e) deleting selected feature pattern, and the remaining pattern features combined into a composite image (f) checking whether the quality of the combined image is distorted; and (g) storing the combined image without distortion in the memory. For example, in the watermarking information removal method described in claim 6, the step (b) of decomposing the watermark image into a corresponding number of feature graphics is implemented by a singular value decomposition algorithm, the singularity The value decomposition algorithm is an algorithm for extracting the watermark information hidden in the watermark image by calculating the maximum singular value point in the watermark image by the quantization strategy. The method for removing watermark information according to claim 6, wherein the step (d) comprises the steps of: calculating a primitive value of each primitive point in the feature graphic; and calculating an average of all the primitive points in the feature graphic. The primitive value is determined whether the primitive value of each primitive point in the feature graphic is greater than the average primitive value; if the primitive value of the primitive point is less than or equal to the average primitive value, the primitive value of the primitive point is converted The binary value is “0”, or if the primitive value of the primitive point is greater than the average primitive value, the primitive value of the primitive point is converted into a binary value “1; each feature graphic is recorded separately The number of feature points whose value is "1", and the number of feature points is added to obtain the total feature value of each feature graphic; the total value of each feature is sorted in order of largest to smallest; and the maximum feature total is selected. The feature graphic corresponding to the value is removed and the feature graphic is removed within the range of the graphic number. The method for removing watermark information according to claim 6, wherein the step (f) includes the following steps: calculating a ratio of a primitive value of the watermark image to a primitive value of the combined image to obtain a peak of the combined image a noise ratio; determining whether the peak noise ratio of the combined image is less than a user-set allowable value; and If the peak noise ratio of the combined image is less than the allowable value set by the user, it indicates that the quality of the combined image has been distorted; or if the peak noise ratio of the combined image is greater than or equal to the allowable value set by the user, it indicates that the combined image is There is no distortion in quality.