CN104915918A - High-robustness digital watermarking method for image maps - Google Patents

Abstract

The invention specifically relates to a high-robustness digital watermarking method for image maps. The high-robustness digital watermarking method comprises the steps of reading a linked list of nodes according a data structure of an IMG image map, and acquiring a pixel value of each node according to a corresponding field; acquiring watermarking binary codes, and generating a corresponding pseudorandom sequence; setting a value which is less than one pixel unit as a threshold value according to the map scale and the size of the pseudorandom sequence, and translating the map in eight directions; calculating a difference value between each feature point and a pixel point around, reading the pseudorandom sequence, and accumulating a product of a corresponding random sequence value and the pixel value difference to the pixel value of the feature point; and randomly selecting several of the maps in which a watermark is embedded to publish. The method provided by the invention realizes high-robustness watermark embedment and extraction. A watermark embedding carrier is based on the brightness difference of images, the watermark is embedded into a plurality of maps with extremely strong correlation, and the watermark can be restored through solving the correlation and an average pixel value of the maps once the watermark is damaged.

Description

A kind of strong robust digital watermark method towards photomap

Technical field

What the present invention was specifically related to is a kind of strong robust digital watermark method towards photomap.

Background technology

Remote sensing image (satellite, aviation, ground close shot) obtains earth surface as earth observation and covers the carrier with structural information, is indispensable information source in geoanalysis application.And how geographical image to be converted into valuable information to successful implementation GIS and drawing engineering be vital.At present, around us, increasing people can utilize FR geographical audio and video products extract and use valuable information.

Raw data as numerous numerical map is originated, authenticity and the security of photomap are especially paid close attention to, and should safeguard the possessory legitimate interests of map, lack map content again and are not tampered, for this reason, photomap digital watermarking research obtains and develops comparatively fast.

At present, little to the digital watermarking algorithm of photomap, the watermarking algorithm for image mainly comprises two classes: i.e. spatial domain and frequency field.Spatial-domain algorithm mainly carrys out embed watermark by amendment Pixel Information, and algorithm realization is simple, but robustness is poor, also comparatively large to the damage of image, cannot resisting various attacks; Frequency field algorithm obtains frequency coefficient by mathematical methods such as discrete cosine transform, Fourier transform or wavelet transformations, embed watermark is carried out by adjustment frequency coefficient, such algorithm is more complicated, also has certain influence to the quality of image, and cannot resist multiple combination attack.

Based on the deficiency of spatial-domain algorithm and frequency field algorithm, some scholars proposed the algorithm of harmless shadow map watermark, mainly will study as follows: Wang Xianmin etc. proposed small echo for the self-adaption two-dimensional blind watermarking algorithm based on characteristics of remote sensing image in 2004; Wang Xun etc. propose a kind of map vector digital watermarking algorithm of robust; Liu Jiufen etc. propose a kind of wavelet transformed domain watermarking algorithm of resist geometric attacks; Zhang Jun etc. have invented the digital watermark based on neural network for image authentication; Fertile spark etc. proposes the Gray-level adaptive blind watermarking of view-based access control model characteristic.This algorithm is the algorithm based on the strong robust watermarking book towards photomap, compared with above algorithm, allows spatial-domain algorithm and frequency field algorithm more optimize.

Summary of the invention

The object of the present invention is to provide one to be highly resistant to various attacks, there is the strong robust digital watermark method towards photomap of high security and invisibility.

The present invention includes following steps:

(1) according to the data structure of IMG photomap, read the chained list of node, and obtain the pixel value of each node according to corresponding field respectively;

(2) scanning watermark bitmap, obtains watermark binary coding, generates corresponding pseudo-random sequence L,

$L=\left\{\begin{array}{c}{\mathrm{Laifm}}_i=1\\ -{\mathrm{Laifm}}_i=0\end{array}\right.;$

Wherein, {+1 ,-1}, L is a pseudo-random sequence be made up of N number of element to La ∈, and N is the length of watermark encoder;

(3) scale and pseudo-random sequence size according to the map, setting is less than the value of 1 pixel unit as threshold value, map is carried out to the translation in eight directions, comprises T, upper right RT, right R, bottom right RD, lower D, lower-left LD, left L, upper left LT;

(4) for eight width maps after displacement, according to clock-wise order, random selecting is 3 width maps wherein, join the atlas { V|v that same original map forms watermark to be embedded jointly _i| 1≤i≤4};

(5) utilize human visual system, detect each width map, according to atlas { V|v _i| 1≤i≤4}, only embeds the coding rule of 1 bit watermark according to each node, choose the unique point set of maximum-norm, extract the pixel value of corresponding node simultaneously;

(6) calculate the difference of each unique point and surrounding pixel point, and read pseudo-random sequence, by the pixel value of the product accumulation of corresponding stochastic ordering train value and pixel value difference to this unique point;

(7) for the map after embed watermark, random selecting wherein a few width is issued;

(8) when watermark detection, by map to be detected and other embed watermark but the map do not issued mutually merge, try to achieve the map with average pixel value;

(9) according to the pseudo-random sequence of original watermark, and the watermark map after merging, compute associations coefficient, if the absolute value of related coefficient is more than or equal to set threshold value, then shows watermark encoder to be detected, otherwise does not detect;

(10) coding is reduced to watermark bitmap, and calculates similarity.

The generation step of the atlas of described watermark to be embedded comprises:

For every width map v _i, calculate corresponding watermark embed strength:

S(v _i)＝{s(p _j)>0|1≤j≤N}

Wherein, s (p _j) represent unique point p in map _jin the intrasystem not discernable degree of HVS, i.e. p _jwith the luminance difference of other eight pixel node around other;

Read watermark sequence successively, to each unique point:

s′(p _j)＝s(p _j)+s(p _j)L _j，

v _i′＝v _i+s(p _j)L _j，

Obtain V ' | v _i' | 1≤i≤4}.

Beneficial effect of the present invention is

The present invention is completely based on the chromatic characteristic of photomap, and the mode coexisted by many figure, is kept the base pixel feature of map, reached the horizontal object of depositing of watermark, specifically comprised:

Achieve watermark embedment and the extraction of strong robustness.Watermark embedment carrier based on the luminance difference of image, and embeds in the extremely strong map of several correlativitys, and watermark, once be destroyed, is recovered by the average pixel value of correlating and map.

Extremely strong robustness.Through test, watermark is seriously damaged, and only needs photomap carrier to have availability, to be then averaged recoverys by correlation map, the watermark information that acquisition similarity is higher.

Good disguise.The selection of watermark embedment node combines HVS principle, unique point is all the position that some visions are not easily discovered, and by calculating luminance difference, effectively control the color disturbance because watermark embedment causes, because the position embedded is very hidden, assailant is difficult to ad hoc launch a offensive to watermark.

Accompanying drawing explanation

Fig. 1 photomap file structure;

Fig. 2 a photomap original map;

Panning effect figure on the right side of Fig. 2 b image ground;

Fig. 2 c image ground bottom right panning effect figure;

Fig. 3 experimental arrangement interface;

Fig. 4 overall system architecture figure;

Fig. 5 watermark embedment process flow diagram.

Embodiment

Below in conjunction with experiment embodiment, the present invention is described in more detail:

What the present invention proposed is a kind of strong robust watermarking method towards photomap.(1) read photomap file, obtain complete data structure; (2) according to the data structure of IMG image, extract map color eigenwert, pixel value is encoded; (3) setting threshold value is combined, by original map according to upper and lower, left and right, upper left, lower-left, upper right, eight, bottom right direction random translation in threshold range; (4) random selecting 4 width from the map that eight width translations obtain, and jointly form image sequence with former figure; (5) make watermark bitmap and be converted to binary sequence, corresponding watermark encoder produces one can for the pseudo-random sequence embedded; (6) to selected image sequence, choose limited unique point in conjunction with HVS (human visual system) principle and threshold value, calculate unique point pixel and the difference of other eight pixels around it; (7) result of product of pixel value difference and pseudo-random sequence is added in unique point accordingly; (8) recurrence completes aforesaid operations, and any a few width in 5 width maps of then random selecting embed watermark releases.

The invention relates to the information processing method of cartography and Geographical Information Sciences, mainly a kind of water mark method of IMG format image map.

The object of the present invention is achieved like this: the embedding and the extraction that comprise watermark, and its step comprises:

(1) read IMG photomap file, obtain the chained list of all Pixel Information.

(2) scanning watermark bitmap, obtains watermark binary coding, and generates pseudo-random sequence.

(3) set threshold value, and in conjunction with threshold value, map is carried out to the translation in eight directions.

(4) obtain eight width maps after displacement, random selecting is three width wherein, join same original map in the lump as the map of watermark to be embedded.

(5) utilize HVS (human visual system) system, detect every width map, obtain unique point.

(6) difference of each unique point and surrounding pixel point is calculated, by the product accumulation of corresponding stochastic ordering train value and pixel value difference to this unique point.

(7) for the map after embed watermark, random selecting wherein a few width is issued.

(8) when watermark detection, by map to be detected and other embed watermark but the map do not issued mutually merge, try to achieve the map with average pixel value.

(9) according to original watermark sequence, and the watermark map after merging, compute associations coefficient.Detect watermark encoder.

(10) coding detected is reduced to watermark bitmap, and calculates the similarity with original watermark.

The present invention can also comprise:

In the step of 2, described reading IMG photomap file, directly according to the data organizational structure of IMG pyramid, from linked list head node, read the pixel value field of each node.

3, in the step that described pseudo-random sequence generates, concrete grammar is:

According to the watermark bitmap defined, be binary watermark encoder sequence by bitmap-converted, then scan each watermark bit value successively, and build pseudo-random sequence L according to following rule.

$L=\left\{\begin{array}{c}{\mathrm{Laifm}}_i=1\\ -{\mathrm{Laifm}}_i=0\end{array}\right.$

Wherein, {+1 ,-1}, then L is a pseudo-random sequence be made up of N number of element to La ∈, and N is the length of corresponding scale-of-two watermark encoder.

4, in described threshold value setting and map translation motion, in order to reduce the strong disturbance excessively for watermark map, limiting translation unit can not more than 1 pixel, therefore threshold value is less than 1, simultaneously, in order to play the object obscured and spread, according to counterclockwise defining eight translation modes, eight width obtaining different directions derive map.

5, described map is chosen in process, it is the relation in order to both balancing safety and efficiency, can not too much embed watermark information frequently, therefore choose 3 width from the derivative map obscured, and in conjunction with original map in the lump as can the map object { V|v of embed watermark _i| 1≤i≤4}.

6, in the process that described every width map feature point is chosen, according to HVS (human visual system) principle, scan each width map, find the key node in the insensitive region of vision, the Selection Strategy of finish node is the rule only embedding 1 bit watermark with reference to every node, chooses maximum element set.

7, in described digital watermark embedding strategy, be added in the insensitive unique point of vision by the product of watermark pseudo-random sequence with luminance difference, concrete grammar is:

For every width map v _i, calculate corresponding watermark embed strength, i.e. watermark capacity:

S(v _i)＝{s(p _j)>0|1≤j≤N}

Wherein, s (p _j) representation feature point p _jkeep the insensitive situation of vision in HVS system under, p _jwith the luminance difference of other eight pixel node around other.

Particularly, after reading the pseudo-random sequence value of watermark, each unique point embed watermark information in map:

s′(p _j)＝s(p _j)+s(p _j)L _j

To every width map recursive calculation thus:

v _i′＝v _i+s(p _j)L _j

Finally obtain the atlas after embed digital watermark V ' | v _i' | 1≤i≤4}.

8, in described watermark map publishing process, namely choose atlas V ' | v _i' | the part map publishing in 1≤i≤4}.

9, in described watermark information testing process, by map V ' to be detected with other embed watermark but the map V do not issued ₀merge mutually, according to threshold value T, reverse translation V ₀, and backward recovery is carried out to map V ' to be detected, try to achieve the map specimen of average pixel value

10, concrete watermark extracting mode is: according to original watermark W ₀create-rule, regenerate pseudo-random sequence L, calculate L and map specimen degree of association C:

$\begin{array}{c}C=\frac{1}{N}\mathrm{\Σ}{L}_{i}S\left(\stackrel{\‾}{V}\right)=\frac{1}{N}\mathrm{\Σ}{L}_i\left(s\right(q_j)+s\left(q_j\right)L_j)\\ =\frac{1}{N}\mathrm{\Σ}L\left(S\left(V_0\right)\right)+\frac{1}{N}\mathrm{\Σ}s\left(q_j\right)\end{array}$

If | C|>T, shows that map has been embedded in doubtful watermark W _t; Otherwise show that pseudo-random sequence is not present in map, i.e. the detected watermark of nothing in map, is specifically expressed as:

$W_t=\left\{\begin{array}{cc}success,& C>T\\ success,& C<-T\\ failed,& -T<C<T\end{array}\right.$

Especially, in unit watermark comparison procedure, be tied to form vertical just like ShiShimonoseki:

$w_i=\left\{\begin{array}{cc}1,& c_i>T\\ 0,& c_i<-T\\ failed,& -T<c_i<T\end{array}\right.$

Wherein, w _i∈ W _t, Σ _ic _i=C.

11, for the watermark sequence of trying to achieve, by the Similarity Measure with original watermark, the authenticity of watermark and the integrality of map is determined.For the watermark identifier W detected _t, cause partial information bit loss because map is under attack, cause W _tinconsistent with former watermark W content.To this, need to carry out similarity watermark detection:

$Sim\left(W_0,W_t\right)=\frac{\underset{i=1}{\overset{n}{\&Sigma;}}<W_i,W_{t_i}>/N}{\sqrt{\underset{i=1}{\overset{n}{\&Sigma;}}{W_i}^2}\&CenterDot;\sqrt{\underset{i=1}{\overset{n}{\&Sigma;}}{W_{t_i}}^2}},<W_i,W_{t_i}>=\left\{\begin{array}{cc}1& ,W_i=W_{t_i}\\ 0& ,W_i\&NotEqual;W_{t_i}\end{array}\right.$

W ₀for original watermark, W _tfor the watermark detected.N is watermark capacity.For there are two kinds of situations: 1. bit numerical value is contrary; 2. the special bit value of body is empty.

Embodiment 1

Towards a strong robust digital watermark method for photomap, comprise embedding and the extraction of watermark, its step comprises:

(1) read IMG photomap file, obtain the chained list of all Pixel Information.

(2) scanning watermark bitmap, obtains watermark binary coding, and generates corresponding pseudo-random sequence.

(3) scale and pseudo-random sequence size according to the map, automatically sets threshold value, and in conjunction with threshold value, map is carried out to the translation in eight directions.

(4) for eight width maps after displacement, random selecting three width wherein, join same original map in the lump as the map of watermark to be embedded.

(5) utilize HVS (human visual system) system, detect each width map, obtain its unique point and pixel value thereof.

(6) calculate the difference of each unique point and surrounding pixel point, and read pseudo-random sequence, by the pixel value of the product accumulation of corresponding stochastic ordering train value and pixel value difference to this unique point.

(7) for the map after embed watermark, random selecting wherein a few width is issued.

(8) when watermark detection, because the map in application process can be subject to change in various degree, for this reason, by map to be detected and other embed watermark but the map do not issued mutually merge, try to achieve the map with average pixel value.

(9) according to the pseudo-random sequence of original watermark, and the watermark map after merging, compute associations coefficient.If the absolute value of related coefficient is more than or equal to set threshold value, then shows watermark encoder to be detected, otherwise do not detect.

(10) by the way, coding is reduced to watermark bitmap, and calculates similarity.

According to the data structure of IMG photomap, read the chained list of node, and obtain the pixel value of each node according to corresponding field respectively.

According to the watermark bitmap defined, be expressed as binary watermark encoder sequence, and built pseudo-random sequence L according to following rule.

$L=\left\{\begin{array}{c}{\mathrm{Laifm}}_i=1\\ -{\mathrm{Laifm}}_i=0\end{array}\right.---\left(1\right)$

Wherein, {+1 ,-1}, L is a pseudo-random sequence be made up of N number of element to La ∈, and N is the length of watermark encoder.

Random selecting value being less than 1 pixel unit as threshold value, and according to threshold size according to clockwise direction respectively translation obtain eight width maps.Direction is: upper (T), upper right (RT), right (R), bottom right (RD), under (D), lower-left (LD), left (L), upper left (LT).

Consider robustness and counting yield, according to clock-wise order, random selecting is 3 width maps wherein, join the atlas { V|v that same original map forms watermark to be embedded jointly _i| 1≤i≤4}.

Gather { V|v according to the map _i| 1≤i≤4}, according to HVS (human visual system) principle, only embeds the coding rule of 1 bit watermark according to each node, choose the unique point set of maximum-norm, extract the pixel value of corresponding node simultaneously.

For every width map v _i, calculate corresponding watermark embed strength:

S(v _i)＝{s(p _j)>0|1≤j≤N} (2)

Wherein, s (p _j) represent unique point p in map _jin the intrasystem not discernable degree of HVS, i.e. p _jwith the luminance difference of other eight pixel node around other.Guarantee that watermark embedment does not cause excessive disturbance to map with this.

Particularly, read watermark sequence successively, to each unique point:

s′(p _j)＝s(p _j)+s(p _j)L _j(3)

Thus:

v _i′＝v _i+s(p _j)L _j(4)

Obtain the atlas after embed watermark V ' | v _i' | 1≤i≤4}.

The watermark extracting that the present invention is conveniently follow-up, can choose arbitrarily 3 width map publishing at the most from V '.

The present invention, according to the watermark map do not issued and map to be detected, according to threshold definitions, after selected part area pixel value contrasts, carries out backward recovery to all translation maps, and finally tries to achieve the map specimen of average pixel value.

The present invention, according to original watermark create-rule, regenerates pseudo-random sequence, and calculate the degree of association of the map specimen of itself and average pixel value, if random series is present in map, then the absolute value of the degree of association must be greater than threshold value; The degree of association is too small, then show that pseudo-random sequence is not present in map, i.e. the detected watermark of nothing in map.

The present invention, for the watermark sequence of trying to achieve, by the Similarity Measure with original watermark, determines the authenticity of watermark and the integrality of map.

The present invention is towards the strong robust digital watermark method of photomap, and carrier is the Digital image map of IMG form, and IMG directly can be converted to the multiple formats such as TIFF, JPG, CNG.Development environment is VC.6.0, as shown in Figure 4, and the embedding grammar of main verifying digital watermark and extracting method.

(1) data structure of photomap file interior joint Pixel Information is read

Organizational form due to the node of IMG file is similar to binary tree (as shown in Figure 1).When reading the node pixel data of IMG file, adopt first order traversal, each node has oneself chained list and header file, and the storage organization EhfaEntry form of header file is as follows:

Longnext; The position * of/* next node/

Longprey; The position * of the previous node of/*/

Longparent; The position * of/* father node/

Longchild; The position * of/* the first child node/

Longdata; / * the deposit position of pixel data*/

Longdatasize :/* size of data */

Char [64] name; The name * of/* node/

Char [32] type; The storage organization * of/* node/

TIMEmodTime; The modification time * of this node of/*/

Concrete Pixel Information is had after the data field of each node.

(2) watermark sequence is produced by scanning input watermark bitmap

Program reads watermark bitmap, and the pixel value of surface sweeping bitmap, obtains binary sequence.In order to improve security, generate corresponding pseudo-random sequence.

(3) unique point of every width map is extracted

As shown in Figure 2, after selecting translation, need the object of embed watermark, by watermark encoder length write HVS system, particularly:

A reads photomap file, according to pseudo-random sequence size, determines the number of unique point;

B utilizes HVS system, scanning full width map, and the insensitive unique point of search maximum quantity, if unique point has adjacencies, then compares its brightness value;

If c alternative numbers is still greater than watermark encoder length, then sort according to brightness value, determine priority;

D preserves unique point set.

(4) pseudo-random sequence is write map successively

Concrete steps are as follows:

A reads the Pixel Information of first Object node.

B compares the luminance difference of this node and surrounding adjacent node, records this difference.

C is process flow diagram as shown in Figure 5, watermark encoder sequence is write in the pixel data of each node successively, by the pixel value of luminance difference (noting just can bearing) product accumulation of encoding corresponding to pseudo-random sequence to this node.

D preserves map file.

As shown in Figure 3, be master routine interface.

(5) obtain new map specimen, extract digital watermarking

Read map file to be detected, according to the map file of identical content, again translation recovery is carried out to map, ask for the average pixel value detecting map, obtain map specimen, and calculate the degree of correlation of map specimen and pseudo-random sequence, and if the absolute value of related coefficient is greater than threshold value, corresponding watermark extracting success.

(6) according to the binary watermarking sequence of reading and the size of watermark bitmap, generating watermark bitmap.

According to the byte number of the watermark bitmap extracted, bitmap byte number=bitmap width * Bitmap Height, and the binary watermarking sequence read can generate original watermark.And authenticity and the integrated degree that NC value relatively determines watermark is carried out to original watermark bitmap and the watermark bitmap detected.

Claims (2)

1., towards a strong robust digital watermark method for photomap, comprise embedding and the extraction of watermark, it is characterized in that, comprise the steps:

(1) according to the data structure of IMG photomap, read the chained list of node, and obtain the pixel value of each node according to corresponding field respectively;

(2) scanning watermark bitmap, obtains watermark binary coding, generates corresponding pseudo-random sequence L,

Wherein, {+1 ,-1}, L is a pseudo-random sequence be made up of N number of element to La ∈, and N is the length of watermark encoder;

(3) scale and pseudo-random sequence size according to the map, setting is less than the value of 1 pixel unit as threshold value, map is carried out to the translation in eight directions, comprises T, upper right RT, right R, bottom right RD, lower D, lower-left LD, left L, upper left LT;

(4) for eight width maps after displacement, according to clock-wise order, random selecting is 3 width maps wherein, join the atlas { V|v that same original map forms watermark to be embedded jointly _i| 1≤i≤4};

(5) utilize human visual system, detect each width map, according to atlas { V|v _i| 1≤i≤4}, only embeds the coding rule of 1 bit watermark according to each node, choose the unique point set of maximum-norm, extract the pixel value of corresponding node simultaneously;

(6) calculate the difference of each unique point and surrounding pixel point, and read pseudo-random sequence, by the pixel value of the product accumulation of corresponding stochastic ordering train value and pixel value difference to this unique point;

(7) for the map after embed watermark, random selecting wherein a few width is issued;

(8) when watermark detection, by map to be detected and other embed watermark but the map do not issued mutually merge, try to achieve the map with average pixel value;

(9) according to the pseudo-random sequence of original watermark, and the watermark map after merging, compute associations coefficient, if the absolute value of related coefficient is more than or equal to set threshold value, then shows watermark encoder to be detected, otherwise does not detect;

(10) coding is reduced to watermark bitmap, and calculates similarity.

2. a kind of strong robust digital watermark method towards photomap according to claim 1, it is characterized in that, the generation step of the atlas of described watermark to be embedded comprises:

For every width map v _i, calculate corresponding watermark embed strength:

S(v _i)＝{s(p _j)>0|1≤j≤N}

Wherein, s (p _j) represent unique point p in map _jin the intrasystem not discernable degree of HVS, i.e. p _jwith the luminance difference of other eight pixel node around other;

Read watermark sequence successively, to each unique point:

s′p _j＝s(p _j)+s(p _j)L _j，

v _i′＝v _i+s(p _j)L _j，

Obtain V ' | v _i' | 1≤i≤4}.