A kind of fragile watermarks technology is applied to the method for digital image fidelity
Technical field
The present invention relates to use computer software, in digital image, embed the summary data of image itself, can detect the change of visual 1bit. " watermark " among the present invention's " a kind of fragile watermarks technology is applied to the method for digital image fidelity " refers to digital watermarking, that relevant with image content or more incoherent beacon information are directly embedded in the middle of the image, but do not affect the use value of raw content, and be not easy to be perceiveed by human visual system or notice. After " fragile watermarks " referred to that image is subject to any processing attack, watermark information was all fully destroyed. " be applied to digital image " and represent that our digital watermark is applied to still image, rather than the multimedia messages such as other audio frequency, video, three-dimensional grid. " method of fidelity " shown the purpose of our digital watermark, and for the true integrated authentication of image content, the watermark of embedding is all responsive to the very small change of image.
Background technology
In recent years, along with the progress of digitizing technique and developing rapidly of internet, the application of digital multimedia (image, video, audio frequency and three dimensional computer graphics etc.) has obtained breakthrough. The advantage that Digital Media is easy to edit, revise, copy and propagate also makes its intellectual property protection and the problems such as authentication of authenticity (reality) and integrality (integrity) become the focus that people pay close attention to when the promotion informationized society is advanced. Since the early 1990s, it is the research field of a focus that digital watermarking and Information hiding develop rapidly, has attracted domestic and international a large amount of research and working people.
The research worker finds that data are hidden and digital watermark technology may have very successful application in the authentication of multi-medium data. At present people to the attention of this problem not enough. The authentication of multi-medium data will be confirmed data whether complete (have or not and alter), true and from the horse's mouth (the alleged source of data) exactly. The authentication of multi-medium data can be used as a kind of electronic evidence, at e-commerce and e-government, in fields such as court evidence and military affairs, medical document and media very exigence is arranged.
The fidelity of digital image is exactly that true integrality to digital image authenticates. We realize by Fragile Watermarking Technique. We with the naked eye can not differentiate their difference the image of watermarked front and back. Can consult following youngster's piece of writing for the digital watermark of Digital image authentication and describe paper.
[1]D.Kundur,and D.Hatzinakos,Digital watermarking for telltale tamper proofing and authentication.
Proc.of the IEEE,87(7),1167-1180,1999
[2]M.Wu,and B.Liu,Watermarking for image authentication.Proc.of ICIP,vol.2,427-431,1998
[3]M.Yeung and F.Mintzer,Invisible watermarking for image verification.J.of Electronic Imaging,7(3),
1998
The digital watermark technology that is used for authentication is normally for gray scale and chromatic image, and our method can be processed the digital image of various bit forms.
Summary of the invention
The present invention's's " a kind of fragile watermarks technology is applied to the method for digital image fidelity " purpose is: disclose a kind of computer processing method that embeds and extract the image content fidelity of data in image. This method is by the numerical characteristic value of digital digest algorithm computational picture content, then numerical characteristic value is embedded in the image itself according to rapid wear pattern formula as digital watermark information and goes, so-called rapid wear pattern formula, just refer to that image is subject to any processing and attacks after, watermark information all can change. For the image before and after the embedding data, we with the naked eye can not tell their difference. With the digital watermarking of this fabrication techniques, all responsive to the change that image is very small, can detect the change of visual 1bit. Therefore can well be used for the authentication of the true integrality of image content, be used as the purpose of content fidelity.
In order to achieve the above object, the present invention is performed such. By the numerical characteristic value of digital digest algorithm extraction image content, then numerical characteristic value is embedded in the image itself according to rapid wear pattern formula as digital watermark information and goes. The numerical characteristic value that recomputates when verifying for the embedding impact that prevents watermark information, we are divided into two parts to the information in the image, and a part is used as the numerical characteristic value that generates image content, and a part is used for hiding watermark information. After image is changed, or be that visual numerical characteristic value has changed, or be that the watermark information of hiding has changed. We relatively these two, if they are identical, key diagram as if truly be not tampered, otherwise just the explanation image had been changed.
Method of the present invention comprises the method that embeds the summary watermark and extracts the method for summary watermark. Embed the method for summary watermark, comprise the embedding of binary picture summary watermark, the embedding of the embedding of 8bit gray scale image summary watermark, the watermark of 8bit chromatic image summary, four submodules of embedding of 24bit chromatic image summary watermark. According to the difference of image to be processed, we are through certain judgement, will enter the make a summary embedding operation of watermark of some modules, will obtain containing at last the image of watermark. Extract the method for summary watermark, comprise the extraction of binary picture summary watermark, the extraction of the extraction of 8bit gray scale image summary watermark, the watermark of 8bit chromatic image summary, four submodules of extraction of 24bit chromatic image summary watermark. According to the difference of image to be processed, we are through certain judgement, with entering the make a summary extraction operation of watermark of some modules, draw at last judged result, finish the checking work to the true integrality of image. Now be described in detail as follows:
The method that embeds the summary watermark comprises following 14 steps:
The first step is to judge first the format information of image:
If image is the 1bit image, carry out second step to the six steps operation;
If image is the 8bit image, judge whether coloured image; To carry out the tenth, ten single steppings; Not to judge whether again it is by the 1bit image transitions; Be, be reduced to the 1bit image, carry out second step to the six steps operation, again the 1bit that obtains is contained the image that watermarking images is converted into 8bit; Not to carry out for the 7th step to the operation of the 9th step;
If the 24bit image judges whether coloured image; To carry out the 12 to the tenth FOUR EASY STEPS; Not that judgement is by the 1bit image or by the 8bit greyscale image transitions; If be converted by the 1bit image, be the 1bit image with image restoring, then carry out second step to the six steps operation, again the 1bit watermarking images that obtains is converted into the image of 24bit; If be converted by the 8bit gray level image, be the 8bit gray level image with image restoring, then carried out for the 7th step to the operation of the 9th step, again the 8bit Gray-level Watermarking image that obtains is converted into the image of 24bit;
Obtain watermark image, termination routine.
Second step is the zone of determining the hiding summary of binary picture, and the zone is the minimum rectangular area that has comprised all stains, and later processing is all carried out in this zone;
The 3rd step was that bianry image is converted to an one-dimension array, if certain pixel is stain in the image, then corresponding array element value is 1 (or 0), if white point, then corresponding array element value is 0 (or 1);
The 4th step was to meet the number that the grid of pattern can be turned in the center in the statistical picture. If number is not less than the figure place of wanting embedding data, then be chosen in the embedding summary data that repeats in these points; If less than the figure place of wanting embedding data, calculate again the number of the grid outside 16 templates of removing definition. Be not less than the figure place of wanting embedding data if remove the number of the outer grid of 16 templates of definition, then in these grids, choose the respective counts point of destination according to scanning sequency with identical interval and come embedding data; If less than, then the number of 4 templates grid outward of definition is removed in calculating. Be not less than the figure place of wanting embedding data if remove the number of the outer grid of 4 templates of definition, then in these grids, choose the respective counts point of destination according to scanning sequency with identical interval and come embedding data; If less than, termination routine returns error message;
The 5th step was the selection according to top embedding point, and the element value that the embedding data of selecting is put in the corresponding one-dimension array is all put 1 (or 0); Then the one-dimension array that obtains is carried out Hash (Hash) conversion with the digital digest algorithm, obtain unique summary info; Utilize key to carry out scramble to the unique digest value that obtains at last;
The 6th step was that the digest value behind the scramble is embedded in the embedding information point of just now selecting. If the data bit of summary is 1, makes that corresponding embedding data point is stain (or white point), otherwise be white point (or stain).
The 7th step was to judge that image has and do not have enough picture elements to embed summary info, if do not have, termination routine returns error code;
The 8th step was that original image is converted to an one-dimension array, and the value of each array element is the result behind last position 0 (or 1) of pixel value of corresponding image pixel; Then the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; Utilize the unique digest value that obtains key to carry out scramble at last;
The 9th step was the LSB of each pixel of embedding of the digest value repetition behind the scramble that will obtain.
The tenth step was that original image is converted to an one-dimension array, the value of each array element of front portion of array is the pixel value of the image pixel of correspondence, array also has 768 elements afterwards, has stored the rgb value of 256 kinds of colouring informations of video palette, wherein the LSB position 0 of B value; Afterwards the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; Utilize key to carry out scramble to the digest value that obtains at last;
The 11 step was with LSB blue in 256 kinds of colouring informations of embedding palette of the repetition of the digest value behind the scramble.
The 12 step was to judge that image has and do not have enough picture elements to embed summary info, if do not have, termination routine returns error code;
The 13 step was to be an one-dimension array with image transitions, and the blueness of each image, green, red brightness value become respectively the element value of 3 array elements; Then the LSB with the element of the blue brightness value of storage in the one-dimension array sets to 0; Afterwards the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; Utilize key to carry out scramble to the digest value that obtains at last;
The 14 step was the LSB of the blue brightness value of each pixel of embedding of the digest value repetition behind the scramble that will obtain.
Extract the method for summary watermark, can regard the operation opposite with the step that embeds the summary watermark as, be divided into following ten five steps and carry out:
The first step is to judge first the format information of image:
If image is the 1bit image, carry out second step to the six steps operation;
If the 8bit image judges whether coloured image; To carry out the operation of the tenth to the 12 step; Not to judge whether again it is by the 1bit image transitions; Be, be reduced to the 1bit image, carry out second step to the six steps operation; Not to carry out for the 7th step to the operation of the 9th step;
If the 24bit image judges whether coloured image; To carry out for the 13 step to the operation of the 15 step; Not that judgement is by the 1bit image or by the 8bit greyscale image transitions; If be converted by the 1bit image, be the 1bit image with image restoring, then carry out second step to the six steps operation; If be converted by the 8bit gray level image, be the 8bit gray level image with image restoring, then carried out for the 7th step to the operation of the 9th step;
Draw judged result, termination routine.
Second step is the zone of at first determining hiding data, and the zone is the minimum rectangular area that has comprised all stains, and later processing is all carried out in this zone; If find that image is full white region, namely having a few all is white point, and the information that image is tampered is directly returned in so EP (end of program);
The 3rd step was that bianry image is converted to an one-dimension array, if in the image be stain, then corresponding array element value is 1 (or 0), if white point, then corresponding array element value is 0 (or 1);
The 4th step was to meet the number that the grid of pattern can be turned in the center in the statistical picture. Embed the figure place that number is broadcast if number is not less than, then be chosen in these points and extract embedding data; If number less than the figure place of embedding data, is calculated the number of the grid outside 16 templates of removing definition again. If the number of removing the outer grid of 16 templates of definition is not less than the figure place of embedding data, then in these grids, chooses the respective counts point of destination according to scanning sequency with identical interval and extract embedding data; If less than, then the number of 4 templates grid outward of definition is removed in calculating. If the number of removing the outer grid of 4 templates of definition is not less than the figure place of embedding data, then in these grids, chooses the respective counts point of destination according to scanning sequency with identical interval and extract embedding data; If less than, termination routine returns error code;
The 5th step was the selection according to top extraction data point, and the element value that the hiding data of selecting is put in the corresponding one-dimension array is all put 1 (or 0); Then the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; Utilize key to carry out scramble to the summary info that obtains again;
The 6th step was to extract embedding data from the extraction data point of just now selecting, and the data of extracting and the scramble that obtains summary are compared, if identical, key diagram similarly is real, otherwise key diagram looks like to be tampered.
The 7th step was to judge that image has and do not have enough picture elements to embed summary info, if do not have, termination routine returns error code;
The 8th step was that original image is converted to an one-dimension array, and the value of each array element is the result behind last position 0 of pixel value of corresponding image pixel; Then the LSB that extracts each pixel obtains watermark information; Afterwards the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; Utilize key to carry out scramble to the summary info that obtains again;
The 9th step was summary info behind the comparison scramble and the watermark information of extraction. If identical, key diagram similarly is real, otherwise key diagram looks like to be tampered.
The tenth step was that original image is converted to an one-dimension array, the value of each array element of front portion of array is the pixel value of the image pixel of correspondence, array also has 768 elements afterwards, has stored the rgb value of 256 kinds of colouring informations of video palette, wherein the LSB position 0 of B value; Afterwards the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; Utilize key to carry out scramble to the digest value that obtains at last;
The 11 step was that LSB blue from 256 kinds of colouring informations of palette of image extracts the watermark information of hiding;
The 12 step was 128 summaries behind the comparison scramble and the watermark information that extracts. If identical, key diagram similarly is real, otherwise key diagram looks like to be tampered.
The 13 step was to judge that image has and do not have enough picture elements to embed summary info, if do not have, termination routine returns error code;
The 14 step was to be an one-dimension array with image transitions, and the blueness of each image, green, red brightness value become respectively the element value of 3 array elements, and then the LSB from the blue brightness value of each pixel extracts the watermark information of hiding; LSB with blue brightness value in the one-dimension array sets to 0 again; At last the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; And utilize key to carry out scramble to the summary info that obtains;
The 15 step was summary info behind the comparison scramble and the watermark information of extraction. If identical, key diagram similarly is real, otherwise key diagram looks like to be tampered.
The present invention's's " a kind of fragile watermarks technology is applied to the method for digital image fidelity " advantage is as follows:
1. because the present invention adopts the digest value of digital digest algorithm computational picture content, according to the uniqueness of digital digest algorithm, guaranteed that the digest value of the image content that calculates is unique. We again are embedded into the digest value that calculates in the image and go, and this has just guaranteed visual any small change, and we can both detect. For the authentication of the true integrality of image provides good assurance. In the time of checking, the watermark information of hiding in the extraction image and the digest value that recomputates image content, relatively the two draws the checking conclusion.
2. the present invention is divided into two parts to the information in the image, and a part is used as the numerical characteristic value that generates image content, and a part is used for hiding watermark information, and this can affect the problem of the numerical characteristic value of image content with regard to the embedding that has solved watermark. After image is changed, or be that visual numerical characteristic value has changed, or be that the watermark information of hiding has changed. We relatively these two, if they are identical, key diagram as if truly be not tampered, otherwise just the explanation image had been changed.
3. the present invention is when embedding summary data, according to the mankind's visual characteristic, success data message is hidden in the most insensitive zone of visual vision. The visual vision difference that guarantees watermarked front and back is minimum.
4. the present invention also utilizes key to carry out the scramble operation to the image content summary data that wish embeds, and has improved to a certain extent confidentiality and the security of algorithm.
Description of drawings
Fig. 1 is flow chart of the present invention;
Fig. 2-the 1st, the common image of 1bit;
Fig. 2-the 2nd, image shown in Fig. 2-1 embeds the image after the summary watermark;
Fig. 3-the 1st, the common gray scale image of 8bit;
Fig. 3-the 2nd, image shown in Fig. 3-1 embeds the image after the summary watermark;
Fig. 4-the 1st, the common color image of 8bit;
Fig. 4-the 2nd, image shown in Fig. 4-1 embeds the image after the summary watermark;
Fig. 5-the 1st, the common color image of 24bit;
Fig. 5-the 2nd, image shown in Fig. 5-1 embeds the image after the summary watermark;
Fig. 6-the 1st, the source color image of 24bit;
Fig. 6-the 2nd, Fig. 6-1 embed the image after the watermark of self making a summary;
Fig. 6-the 3rd, the result of Fig. 6-2;
Fig. 6-the 4th alters the image behind Fig. 6-2;
Fig. 6-the 5th, the result of Fig. 6-4;
Fig. 7 is the Reasonable Mode collection that the bianry image center pixel can turn over;
Fig. 8 be 16 definition can not embedding data point template;
Fig. 9 be 4 definition can not embedding data point template.
The specific embodiment
The present invention is described in further detail below in conjunction with accompanying drawing, and see also accompanying drawing. The operation that embeds the summary watermark among Fig. 1 can be divided into 14 steps, extracts the operation of summary watermark and can regard opposite process as, and totally ten five steps see Table 1 and table 2.
The step that embeds the summary water mark method is as follows: at first, by primary control program, which processing module the control image should enter, and, judges the format information of image that is: if image is the 1bit image, carry out second step to the six steps operation; If image is the 8bit image, judge whether coloured image; To carry out the tenth, ten single steppings; Not to judge whether again it is by the 1bit image transitions; Be, be reduced to the 1bit image, carry out second step to the six steps operation, again the 1bit that obtains is contained the image that watermarking images is converted into 8bit; Not to carry out for the 7th step to the operation of the 9th step; If the 24bit image judges whether coloured image; To carry out the 12 to the tenth FOUR EASY STEPS; Not that judgement is by the 1bit image or by the 8bit greyscale image transitions; If be converted by the 1bit image, be the 1bit image with image restoring, then carry out second step to the six steps operation, again the 1bit watermarking images that obtains is converted into the image of 24bit; If transform rice by the 8bit gray level image, be the 8bit gray level image with image restoring, then carried out for the 7th step to the operation of the 9th step, again the 8bit Gray-level Watermarking image that obtains is converted into the image of 24bit; Obtain watermark image, termination routine; Second step in order to prevent the result of asemantic visual white edge regional effect algorithm, needs to determine the zone of the hiding summary of binary picture, and the zone is the minimum rectangular area that has comprised all stains, and later processing is all carried out in this zone; The 3rd step is for the digital digest algorithm provides input item, in order to obtain the digest value of image content, bianry image need to be converted to an one-dimension array, if certain pixel is stain in the image, then corresponding array element value is 1, if white point, then corresponding array element value is 0; In the 4th step, in order determine to want the position of embedding data point, need to come embedding data according to the point of the algorithm picks some of bianry image fragile watermark; In the 5th step, in order to obtain the summary info of image content, need to according to the selection of top embedding point, all put 1 to the element value that the embedding data of selecting is put in the corresponding one-dimension array; Then the one-dimension array that obtains is carried out Hash (Hash) conversion with the digital digest algorithm, obtain unique summary info; Utilize key to carry out scramble to the unique digest value that obtains at last; The 6th step, carry out the embedding of summary info, the digest value behind the scramble is embedded in the embedding information point of just now selecting. If the data bit of summary is 1, makes that corresponding embedding data point is stain, otherwise be white point; In the 7th step, the judgement image has does not have enough picture elements to embed summary info, if do not have, termination routine returns error code, thereby can judgement 8bit gray scale image hide summary info; The 8th step in order to obtain the summary info of 8bit gray scale image content, need to be converted to original image an one-dimension array, and the value of each array element is the result behind last position 0 of pixel value of corresponding image pixel; Then the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; Utilize the unique digest value that obtains key to carry out scramble at last; In the 9th step, carry out the embedding of data, the LSB of each pixel of embedding that the digest value behind the scramble that obtains is repeated; The tenth step, in order to obtain the summary info of 8bit color graphics, original image need to be converted to an one-dimension array, the value of each array element of front portion of array is the pixel value of the image pixel of correspondence, array also has 768 elements afterwards, stored the rgb value of 256 kinds of colouring informations of video palette, wherein the LSB position 0 of B value; Afterwards the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; Utilize key to carry out scramble to the digest value that obtains at last; The 11 step, in the palette that summary data is embedded into image, blue LSB in 256 kinds of colouring informations of embedding palette that the digest value behind the scramble need to be repeated; In the 12 step, the judgement image has does not have enough picture elements to embed summary info, if do not have, termination routine returns error code; Thereby can judgement 24bit chromatic image hide summary info; The 13 step in order to obtain the summary data of 24bit chromatic image, need to be an one-dimension array with image transitions, and the blueness of each image, green, red brightness value become respectively the element value of 3 array elements; Then the LSB with the element of the blue brightness value of storage in the one-dimension array sets to 0; Afterwards the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; Utilize key to carry out scramble to the digest value that obtains at last; In the 14 step, the LSB of the blue brightness value of each pixel of embedding that repeats by the digest value behind the scramble that will obtain realizes the embedding of the summary info of image.
The step of extracting the summary water mark method is as follows: the first step, carry out primary control program, and which processing module the control image should enter, and, judges the format information of image that is: if image is the 1bit image, carry out second step to the six steps operation; If the 8bit image judges whether coloured image; To carry out for the tenth step to the operation of the 12 step; Not to judge whether again it is by the 1bit image transitions; Be, be reduced to the 1bit image, carry out second step to the six steps operation; Not to carry out for the 7th step to the operation of the 9th step; If the 24bit image judges whether coloured image; To carry out for the 13 step to the operation of the 15 step; Not that judgement is by the 1bit image or by the 8bit greyscale image transitions; If be converted by the 1bit image, be the 1bit image with image restoring, then carry out second step to the six steps operation; If be converted by the 8bit gray level image, be the 8bit gray level image with image restoring, then carried out for the 7th step to the operation of the 9th step; Draw judged result, termination routine; Second step in order to prevent the result of asemantic visual white edge regional effect algorithm, needs to determine the zone of hiding data, and the zone is the minimum rectangular area that has comprised all inns by brigands, and later processing is all carried out in this zone; If finding image is full white region, namely be a little white point, the information that image is tampered is directly returned in so EP (end of program); The 3rd step was converted to an one-dimension array with bianry image, if in the image be stain, then corresponding array element value is 1, if white point, then corresponding array element value is 0; Thereby for the digital digest algorithm provides input item, in order to obtain the digest value of image content; The 4th step, extract watermark information according to the point of the algorithm picks some of bianry image fragile watermark, determine to want the position of the point of embedding data; In the 5th step, according to the selection of top extraction data point, the element value that the hiding data of selecting is put in the corresponding one-dimension array is all put 1; Then the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; Utilize key to carry out scramble to the summary info that obtains again; Thereby obtain the digest value of binary picture content; The 6th step, in order to obtain being hidden in the digest value in the image, compare the two, draw the checking conclusion, need to extract embedding data from the extraction data point of just now selecting, the data of extracting and the scramble that obtains summary are compared, if identical, key diagram similarly is real, otherwise key diagram looks like to be tampered; In the 7th step, the judgement image has does not have enough picture elements to embed summary info, if do not have, termination routine returns error code; Thereby can judgement 8bit gray scale image hide summary info; The 8th step for the summary info that obtains image content and the watermark summary info of hiding, need to be converted to original image an one-dimension array, and the value of each array element is the result behind last position 0 of pixel value of corresponding image pixel; Then the LSB that extracts each pixel obtains watermark information; Afterwards the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; Utilize key to carry out scramble to the summary info that obtains again; The 9th step in order to compare the two, thereby drew the checking conclusion, needed summary info behind the scramble relatively and the watermark information of extraction. If identical, key diagram similarly is real, otherwise key diagram looks like to be tampered; The tenth step, in order to obtain the summary info of 8bit chromatic image content, original image need to be converted to an one-dimension array, the value of each array element of front portion of array is the pixel value of the image pixel of correspondence, array also has 768 elements afterwards, stored the rgb value of 256 kinds of colouring informations of video palette, wherein the LSB position 0 of B value; Afterwards the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; Utilize key to carry out scramble to the digest value that obtains at last; In the 11 step, LSB blue from 256 kinds of colouring informations of palette of image extracts the watermark information of hiding, thereby obtains being hidden in the watermark information in the palette; The 12 step in order to compare the two, drew the checking conclusion, relatively 128 summaries behind the scramble and the watermark informations that extract; If identical, key diagram as if real, otherwise, illustrate that image was tampered; In the 13 step, the judgement image has does not have enough picture elements to embed summary info, if do not have, termination routine returns error code; Thereby can judgement 24bit chromatic image hide summary info; The 14 step, in order to obtain the summary info of 24bit chromatic image content, obtain being hidden in the summary info in the watermark, need to be an one-dimension array with image transitions, the blueness of each image, green, red brightness value become respectively the element value of 3 array elements, and then the LSB from the blue brightness value of each pixel extracts the watermark information of hiding; LSB with blue brightness value in the one-dimension array sets to 0 again; At last the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; And utilize key to carry out scramble to the summary info that obtains; The 15 step, the relatively summary info behind the scramble and the watermark information of extraction. If identical, key diagram similarly is real, otherwise key diagram looks like to be tampered; Namely compare the two, draw the checking conclusion.
Idiographic flow is referring to subordinate list 1 and subordinate list 2.
Table 1 embeds the operating procedure of summary watermark
Step | Operation | Effect |
1
| Judge the format information of image: if image is the 1bit image, carry out second step to the six steps operation; If image is the 8bit image, judge whether coloured image; To carry out the tenth, ten single steppings; Not to judge whether again it is by the 1bit image transitions; Be, be reduced to the 1bit image, carry out second step to the six steps operation, again the 1bit that obtains is contained the image that watermarking images is converted into 8bit; Not to carry out for the 7th step to the operation of the 9th step; If the 24bit image judges whether coloured image; To carry out the 12 to the tenth FOUR EASY STEPS; Not that judgement is by the 1bit image or by the 8bit greyscale image transitions; If be converted by the 1bit image, be the 1bit image with image restoring, then carry out second step to the six steps operation, again the 1bit watermarking images that obtains is converted into the image of 24bit; If transform rice by the 8bit gray level image, be the 8bit gray level image with image restoring, then carried out for the 7th step to the operation of the 9th step, again the 8bit Gray-level Watermarking image that obtains is converted into the image of 24bit; Obtain watermark image, termination routine. | Primary control program, which processing module the control image should enter |
2
| Determine the zone of the hiding summary of binary picture, the zone is the minimum rectangular area that has comprised all stains, and later processing is all carried out in this zone; | Prevent the result of asemantic visual white edge regional effect algorithm |
3
| Bianry image is converted to an one-dimension array, if certain pixel is stain in the image, then corresponding array element value is 1, if white point, then corresponding array element value is 0; | For the digital digest algorithm provides input item, in order to obtain the digest value of image content |
4
| Meet the center in the statistical picture and can turn over the number of the grid of pattern. If number is not less than the figure place of wanting embedding data, then be chosen in the embedding summary data that repeats in these points; If number less than the figure place of wanting embedding data, is calculated the number of the grid outside 16 templates of removing definition again. Be not less than the wish embedding if remove the number of the outer grid of 16 templates of definition | Determine to want the position of embedding data point |
| The figure place of data is then chosen the respective counts point of destination according to scanning sequency with identical interval and is come embedding data in these grids; If less than, then the number of 4 templates grid outward of definition is removed in calculating. Be not less than the figure place of wanting embedding data if remove the number of the outer grid of 4 templates of definition, then in these grids, choose the respective counts point of destination according to scanning sequency with identical interval and come embedding data; If less than, termination routine returns error message; | |
5
| According to the selection of top embedding point, the element value that the embedding data of selecting is put in the corresponding one-dimension array is all put 1; Then the one-dimension array that obtains is carried out Hash (Hash) conversion with the digital digest algorithm, obtain unique summary info; Utilize key to carry out scramble to the unique digest value that obtains at last; | Obtain the summary info of image content |
6
| Digest value behind the scramble is embedded in the embedding information point of just now selecting. If the data bit of summary is 1, makes that corresponding embedding data point is stain, otherwise be white point. | Embed summary info |
7
| The judgement image has does not have enough picture elements to embed summary info, if do not have, termination routine returns error code; | Can judgement 8bit gray scale image hide summary info |
8
| Original image is converted to an one-dimension array, and the value of each array element is the result behind last position 0 of pixel value of corresponding image pixel; Then the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; Utilize the unique digest value that obtains key to carry out scramble at last; | Obtain the summary info of 8bit gray scale image content |
9
| The LSB of each pixel of embedding that the digest value behind the scramble that obtains is repeated. | Embedding data |
10
| Original image is converted to an one-dimension array, the value of each array element of front portion of array is the pixel value of the image pixel of correspondence, array also has 768 elements afterwards, has stored the RGB value of 256 kinds of colouring informations of video palette, wherein the LSB position 0 of B value; Afterwards the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; Utilize key to carry out scramble to the digest value that obtains at last; | Obtain the summary info of 8bit chromatic image |
11
| In 256 kinds of colouring informations of embedding palette with the repetition of the digest value behind the scramble | Summary data is embedded into |
| Blue LSB. | In the palette of image |
12
| The judgement image has does not have enough picture elements to embed summary info, if do not have, termination routine returns error code; | Can judgement 24bit chromatic image hide summary info |
13
| Be an one-dimension array with image transitions, the blueness of each image, green, red brightness value become respectively the element value of 3 array elements; Then the LSB with the element of the blue brightness value of storage in the one-dimension array sets to 0; Afterwards the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; Utilize key to carry out scramble to the digest value that obtains at last; | Obtain the summary data of 24bit chromatic image |
14
| The LSB of the blue brightness value of each pixel of embedding that the digest value behind the scramble that obtains is repeated. | Embed the summary info of image |
Table 2 extracts the operating procedure of summary watermark
Step | Operation | Effect |
1
| Judge the format information of image: if image is the 1bit image, carry out second step to the six steps operation; If the 8bit image judges whether coloured image; To carry out for the tenth step to the operation of the 12 step; Not to judge whether again it is by the 1bit image transitions; Be, be reduced to the 1bit image, carry out second step to the six steps operation; Not to carry out for the 7th step to the operation of the 9th step; If the 24bit image judges whether coloured image; To carry out for the 13 step to the operation of the 15 step; Not that judgement is by the 1bit image or by the 8bit greyscale image transitions; If be converted by the 1bit image, be the 1bit image with image restoring, then carry out second step to the six steps operation; If be converted by the 8bit gray level image, be the 8bit gray level image with image restoring, then carried out for the 7th step to the operation of the 9th step; Draw judged result, termination routine. | Primary control program, which processing module the control image should enter |
2
| Determine the zone of hiding data, the zone is the minimum rectangular area that has comprised all stains, and later processing is all carried out in this zone; | Prevent the result of asemantic visual white edge regional effect algorithm |
| If finding image is full white region, namely be a little white point, the information that image is tampered is directly returned in so EP (end of program); | |
3
| Bianry image is converted to an one-dimension array, if in the image be stain, then corresponding array element value is 1, if white point, then corresponding array element value is 0; | For the digital digest algorithm provides input item, in order to obtain the digest value of image content |
4
| Meet the center in the statistical picture and can turn over the number of the grid of pattern. If number is not less than the figure place of embedding data, then is chosen in these points and extracts embedding data; If number less than the figure place of embedding data, is calculated the number of the grid outside 16 templates of removing definition again. If the number of removing the outer grid of 16 templates of definition is not less than the figure place of embedding data, then in these grids, chooses the respective counts point of destination according to scanning sequency with identical interval and extract embedding data; If less than, then the number of 4 templates grid outward of definition is removed in calculating. If the number of removing the outer grid of 4 templates of definition is not less than the figure place of embedding data, then in these grids, chooses the respective counts point of destination according to scanning sequency with identical interval and extract embedding data; If less than, termination routine returns error code; | Determine the position of the point of wish embedding data |
5
| According to the selection of top extraction data point, the element value that the hiding data of selecting is put in the corresponding one-dimension array is all put 1; Then the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; Utilize key to carry out scramble to the summary info that obtains again; | Obtain the digest value of binary picture content |
6
| Extract embedding data from the extraction data point of just now selecting, the data of extracting and the scramble that obtains summary are compared, if identical, key diagram similarly is real, otherwise key diagram looks like to be tampered. | Obtain being hidden in the digest value in the image, compare the two, draw the checking conclusion |
7
| The judgement image has does not have enough picture elements to embed summary info, if do not have, termination routine returns error code; | Can judgement 8bit gray scale image hide summary info |
8
| Original image is converted to an one-dimension array, and the value of each array element is the knot behind last position 0 of pixel value of corresponding image pixel | Obtain the summary info of image content and the watermark summary info of hiding |
| Really; Then the LSB that extracts each pixel obtains watermark information; Afterwards the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; Utilize key to carry out scramble to the summary info that obtains again; | |
9
| Summary info behind the comparison scramble and the watermark information of extraction. If identical, key diagram similarly is real, otherwise key diagram looks like to be tampered. | Compare the two, draw the checking conclusion |
10
| Original image is converted to an one-dimension array, the value of each array element of front portion of array is the pixel value of the image pixel of correspondence, array also has 768 elements afterwards, has stored the rgb value of 256 kinds of colouring informations of video palette, wherein the LSB position 0 of B value; Afterwards the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; Utilize key to carry out scramble to the digest value that obtains at last; | Obtain the summary info of 8bit chromatic image content |
11
| LSB blue from 256 kinds of colouring informations of palette of image extracts the watermark information of hiding; | Obtain being hidden in the watermark information in the palette |
12
| 128 summaries behind the comparison scramble and the watermark informations that extract. If identical, key diagram similarly is real, otherwise key diagram looks like to be tampered. | Compare the two, draw the checking conclusion |
13
| The judgement image has does not have enough picture elements to embed summary info, if do not have, termination routine returns error code; | Can judgement 24bit chromatic image hide summary info |
14
| Be an one-dimension array with image transitions, the blueness of each image, green, red brightness value become respectively the element value of 3 array elements, and then the LSB from the blue brightness value of each pixel extracts the watermark information of hiding; LSB with blue brightness value in the one-dimension array sets to 0 again; At last the one-dimension array that obtains is carried out hash conversion with the digital digest algorithm, obtain unique summary info; And utilize key to carry out scramble to the summary info that obtains; | Obtain the summary info of 24bit chromatic image content, obtain being hidden in the summary info in the watermark |
15
| Summary info behind the comparison scramble and the watermark information of extraction. If identical, key diagram similarly is real, otherwise key diagram looks like to be tampered. | Compare the two, draw the checking conclusion |
Please see Fig. 2-1 and Fig. 2-2 in the accompanying drawing. Wherein Fig. 2-1 is the two-value caricature image of a width of cloth 1bit, and content is Ai Erre " sound of chopping wood Adventures of Pei ". Fig. 2-the 2nd, image shown in Fig. 2-1 embeds the image after the summary watermark. The digital digest algorithm picks of computational picture synopsis be the MD5 algorithm, by the MD5 algorithm, can obtain 128 unique summary datas, represent with 16 systems, be " 436740F130AA64800C4C0C735950E5BB ". We are hidden in these 128 summary datas in the image.
Fig. 3-1 and Fig. 3 the-the 2nd, the example of 8bit gray scale image. Fig. 3-1 is a common 8bit gray scale image, Fig. 3-the 2nd, and image shown in Fig. 3-1 embeds the image after the summary watermark. The digital digest algorithm picks of computational picture synopsis be the MD5 algorithm, by the MD5 algorithm, can obtain 128 unique summary datas, represent with 16 systems, be " 5B71C6905978BDA59619B7F45B4F399D ". We are hidden in these 128 summary datas in the image.
Fig. 4-1 and Fig. 4 the-the 2nd, the example of 8bit chromatic image. Fig. 4-1 is a common 8bit chromatic image, Fig. 4-the 2nd, and image shown in Fig. 4-1 embeds the image after the summary watermark. The digital digest algorithm picks of computational picture synopsis be the MD5 algorithm, by the MD5 algorithm, can obtain 128 unique summary datas, represent with 16 systems, be " 8EDF42F48035CD201EA4B2EF347B53A9 ". We are hidden in these 128 summary datas in the palette of image.
Fig. 5-1 and Fig. 5 the-the 2nd, the example of 24bit chromatic image. Fig. 5-1 is a common 24bit chromatic image, Fig. 5-the 2nd, and image shown in Fig. 5-1 embeds the image after the summary watermark. The digital digest algorithm picks of computational picture synopsis be the MD5 algorithm, by the MD5 algorithm, can obtain 128 unique summary datas, represent with 16 systems, be " 9A975353A713BD35A49D6A30DBEFE488 ". We are hidden in these 128 summary datas in the image.
Fig. 6 in the accompanying drawing-1 is an application example of visual fidelity to 6-5. Fig. 6-1 is a common 24bit chromatic image. Fig. 6-the 2nd, image shown in Fig. 6-1 embeds the image after the summary watermark. The digital digest algorithm picks of computational picture synopsis be the MD5 algorithm, by the MD5 algorithm, can obtain 128 unique summary datas, represent with 16 systems, be " 41C9E442EF33C028633576F5E2AF2830 ". We are hidden in these 128 summary datas in the image. Fig. 6-the 3rd, the result of proof diagram 6-2, in the time of checking, utilize on the one hand the MD5 algorithm to recomputate the synopsis of image, represent with 16 systems, be " 41C9E442EF33C028633576F5E2AF2830 ", extract on the one hand the watermark information that is hidden in the image, represent with 16 systems, be " 41C9E442EF33C028633576F5E2AF2830 ", compare the two, be worth identical, key diagram as if truly be not modified. Fig. 6-the 4th alters the later image of Fig. 6-2, and we have made " 1,998 5 30 " to " 1,997 6 30 " in the original picture lower right corner into. We know that the edit-modify of digital image is very simple and easy to do, realize easily. So its true integrality is a problem that people extremely are concerned about. Fig. 6-the 5th, the result of proof diagram 6-4. In the time of checking, utilize on the one hand the MD5 algorithm to recomputate the synopsis of image, represent with 16 systems, be " 008015E8F2CA93D4536D0457226E2619 ", extract on the one hand the watermark information that is hidden in the image, represent with 16 systems, be " E5442CC3B4281EAAAAFBAD00BECAB7C9 " that relatively the two is worth not identical, key diagram likes and is modified, thereby has finished the authentication of visual true integrality.
Fig. 7 is the Reasonable Mode collection that the bianry image center pixel can turn over. If certain grid of 3 * 3 is the some patterns in this set in the image, so the center pixel of this grid be exactly can embedding data point.
Fig. 8 be 16 definition can not embedding data point template. If meet the inadequate embedding data of number of the point of the Reasonable Mode that center pixel can turn over, but just attempt to seek the embedding data point outside the template of embedding data point what remove these 16 definition.
Fig. 9 be 4 definition can not embedding data point template. If remove 16 definition can not embedding data point the inadequate embedding data of number of point of template, but just attempt can not seek the embedding data point outside the template of embedding data point what remove these 4 definition.
So far, be used for embedding the method that the fragile type digital watermarking realizes the digital image fidelity thereby described among the present invention, and from digital image, extract the summary watermark, like no method through revising with proof diagram. For those of ordinary skills, in the situation that does not break away from the spirit and scope that claims of the present invention limit, can make various changes and distortion.