CN103533457A - Multimedia digital watermark embedding and separation methods and multimedia digital watermark embedding and separation devices - Google Patents

Abstract

The invention provides multimedia digital watermark embedding and separation methods and multimedia digital watermark embedding and separation devices. The watermark embedding method comprises the following steps: a presetting stage, a watermark processing stage and a cipher text processing stage, wherein the watermark processing stage comprises steps of calculating watermark fusion information and deciphering the information. According to the scheme, the characteristic of watermark embedding can be set according to the user demand, the robustness is strong, and unique cipher text information capable of analyzing a watermark can be generated for interactive authorization.

Description

Multimedia digital watermark embedding and separation method and device

Technical field

The present invention relates to digital watermark technology field, relate in particular to a kind of multimedia digital watermark embedding and separation method and device.

Background technology

Development along with multimedia technology and social networks; increasing media file is produced and shares on network; Copyright Protection just becomes the problem being more and more concerned; a lot of technology are developed for doing the copyright protection of digital product, and wherein digital watermarking is exactly a kind of more effective protected mode.Multimedia digital watermark, is that copyright information is embedded in digital carrier, neither affects the normal use of original vector, is also not easy to be tampered, thereby reaches the object of copyright protection.In digital living, there are a lot of common multimedia watermark examples, if the TV station's station symbol in television program is exactly a kind of watermark form; In the illustration of Internet news report, embed the sign of source of news; The online video display that Ge great video portal website plays also can be stamped its corresponding commercial mark, to inform copyright owner's identity.

At present, embed watermark information in multimedia file, mainly to adopt in two ways, a kind of is that initial data is processed, namely in the process of media production, the direct embedded images of watermark or frame of video is inner, telescopiny adopts cryptographic means, makes beholder direct viewing to stamp the content of watermark, or after obtaining corresponding deciphering means, watermark can be peeled off and watch the content that no-watermark effect is disturbed; Another kind of mode is directly to add the compression bit stream of data to transmit watermark, has to limit the size of watermarked data volume but this mode is compressed the restriction of bit rate, can not put up with user-defined watermark information.The application uses the former mode, i.e. embed digital watermark in the initial data of media file.

In media, embed the watermark of resembling trade mark, can declare current content all rights reserved information to beholder, who whom by, made by issue etc., being generally making side merges original image or frame of video and watermark information, make the effect of watermark can cover the initial data of certain area, realize visible or hide, recipient extracts the watermark information of embedding by the reverse method merging and recovers the result of use of original contents from corresponding region.If side property in watermark blending algorithm Shandong is strong not, easily by some edit, is carried out Fuzzy processing even directly extract watermark information and distort, thereby may cause infringement of copyright behavior.

The form that watermark presents is varied, from static to motion, from high-visible to hiding to deeply.Wherein, the complexity of making video watermark is higher than still image watermark, and due to continuity rule between frame of video, uses interframe operation method easily the digital watermarking based on original data processing to be extracted, and copyright protection means were lost efficacy.And after identical watermark information incorporates initial data with different definition, through processes such as compressed encoding, transmission and decodings, partial information may be lost, and causes copyright effect to weaken.

Software is made in most of digital watermarking on the market at present, only provides comparatively single static watermark to embed effect.Even if some software can provide some watermark dynamic effects, but only consider the manufacturing process that watermark embeds, do not support watermark separated.

Summary of the invention

The present invention proposes a kind of multimedia digital watermark embedding method, can the feature that watermark embeds be set according to user's request, watermark directly incorporates in original image or video requency frame data, strong robustness, be difficult for being intercepted even and distorting by illegal means, unique cipher-text information that after embedding, generation can be resolved watermark is for mutual mandate.

The present invention also proposes a kind of multimedia digital watermark separation method, can realize the separation of multimedia digital watermark.

Technical scheme of the present invention is achieved in that

A watermark embedding method, comprising: default stage, watermark processing stage and Dealing with encrypt code stage;

The described default stage comprises: import watermark and embed demand; Described watermark embedding demand comprises watermark effect parameter; Described watermark effect parameter comprises video dynamic watermark, watermark original position, dynamic trajectory, movement rate or visible level;

The described watermark processing stage comprises the step of calculating watermark fuse information and decryption information; The step of described calculating watermark fuse information and decryption information comprises:

Calculate key frame watermark and embed region, extract RGBSan road observational components; Use FastICA algorithm to calculate corresponding RGB tri-road separation matrixes; Calculate corresponding RGB tri-tunnel information source arrays; Find the low frequency vector index in each road information source matrix; According to watermark, embed the requirement of demand to visible level, generate and merge matrix; Watermark information RGB component correspondence is mixed into each road RGB low frequency vector of key frame and calculates corresponding RGB tri-road low frequency vectors and the mixed three less important dope vectors in tunnel of watermark; Described RGB tri-road separation matrixes and low frequency vector index are decryption information, and described fusion matrix and less important dope vector are watermark fuse information;

The ciphertext in described Dealing with encrypt code stage comprises: watermark effect parameter, watermark key frame index, decryption information and watermark fuse information.

In said method, watermark embeds demand and can also comprise: the video of local coding and decoding scheme, watermark to be embedded or image, watermark picture and ciphertext encrypted ones.

In said method, described watermark processing is in the stage, and the step of calculating watermark fuse information and decryption information can also comprise before: start encoding and decoding streamline, detect key frame; After the step of described calculating watermark fuse information and decryption information, can also comprise: start watermark and embed streamline, generate watermarked media file.

In said method, described encoding and decoding streamline can be for opening decoding or coding work sequence;

Described decoding work sequence can be: resolve media file, reading out data is also exported the frame data of RGB pattern;

Described coding work sequence can be: receive the frame data of RGB pattern, write in the corresponding media file generating.

For video file, detection key frame can be: the comparatively uniform frame of video of detected energy distribution in video-frequency band, using this frame of video as key frame; For image file, detection key frame can be: directly using described image as key frame.

In said method, starting watermark embedding streamline can comprise the following steps:

(1) calculate present frame watermark and embed region;

(2) resolve present frame watermark and embed region, extract RGBSan road observational components;

(3) use RGBSan road separation matrix to calculate RGBSan road information source matrix array;

(4) utilize fusion matrix that information source low frequency vector and watermark vector are merged mutually;

(5) the main information vector after merging is replaced to original low frequency vector, produce the information source after watermarked;

(6) by separated inverse matrix and watermarked after information source calculate the watermarked RGB observational components in three roads.

In said method, the Dealing with encrypt code stage can comprise: watermark fuse information and decryption information are write to cryptograph files and use password encryption ciphertext.

The present invention also proposes a kind of multimedia digital watermark separation method, comprising: default stage, Dealing with encrypt code stage and watermark processing stage;

The described watermark processing stage comprises: locator key frame, utilize watermark fuse information and decryption information to recover watermark picture;

For image file, described locator key frame is: by key frame index, find key frame position from video sequence; For image file, described locator key frame is: directly using image as key frame;

Described watermark fuse information and the decryption information recovery watermark picture of utilizing comprises the following steps:

Calculate key frame watermark and embed region, extract RGBSan road observational components; Utilize corresponding RGB tri-road separation matrixes to calculate watermarked information source array, and utilize three tunnel information source medium and low frequency vector index to remove to extract three tunnel information source low frequency vectors; The less important dope vector in output San road after utilizing watermark fusion matrix and key frame RGB tri-road low frequency vectors to mix with watermark, recovers watermark information; Wherein, the index of described RGB tri-road separation matrixes and low frequency vector is decryption information, and described fusion matrix and less important dope vector are watermark fuse information.

In said method, the default stage can comprise: import local coding and decoding scheme, watermarked image or video, encryption ciphertext, decrypt ciphertext password.

In said method, the Dealing with encrypt code stage can comprise the information of using in password decrypting ciphertext, parsing ciphertext.

In said method, the watermark processing stage can also comprise: start encoding and decoding streamline, start watermark separated flow waterline and recover original media file.

In said method, starting watermark separated flow waterline can comprise the following steps:

(1) calculate present frame watermark and embed region;

(2) resolve present frame watermark and embed region, extract RGBSan road observational components;

(3) utilize corresponding RGB tri-road separation matrixes to calculate watermarked information source array, and utilize three tunnel information source medium and low frequency vector index to remove to extract three tunnel information source low frequency vectors;

(4) utilize watermark to merge matrix and watermark Vector Message, calculate the three original low frequency vectors in tunnel;

(5) utilize separation matrix inverse matrix to recover the original observation array information in RGBSan road.

A watermark flush mounting, comprising:

User interface is abstract, for importing watermark, embeds demand; Described watermark embedding demand comprises watermark effect parameter; Described watermark effect parameter comprises video dynamic watermark, watermark original position, dynamic trajectory, movement rate or visible level;

Watermark processing engine, for calculating watermark fuse information and decryption information; The step of described calculating watermark fuse information and decryption information comprises: calculate key frame watermark and embed region, extract RGBSan road observational components; Use FastICA algorithm to calculate corresponding RGB tri-road separation matrixes; Calculate corresponding RGB tri-tunnel information source arrays; Find the low frequency vector index in each road information source matrix; According to watermark, embed the requirement of demand to visible level, generate and merge matrix; Watermark information RGB component correspondence is mixed into each road RGB low frequency vector of key frame and calculates corresponding RGB tri-road low frequency vectors and the mixed three less important dope vectors in tunnel of watermark; Described RGB tri-road separation matrixes and low frequency vector index are decryption information, and described fusion matrix and less important dope vector are watermark fuse information;

Encrypt ciphertext module, for carrying out Dealing with encrypt code, the ciphertext of described Dealing with encrypt code comprises: watermark effect parameter, watermark key frame index, decryption information and watermark fuse information.

Said apparatus can also comprise:

Coding/decoding module, for opening decoding or coding work sequence;

Described decoding work sequence is: resolve media file, reading out data is also exported the frame data of RGB pattern;

Described coding work sequence is: receive the frame data of RGB pattern, write in the corresponding media file generating.

A watermark separator, comprising:

User interface is abstract, default for carrying out;

Encrypt ciphertext module, for carrying out Dealing with encrypt code;

Watermark processing engine, for carrying out watermark processing, comprising: locator key frame, utilize watermark fuse information and decryption information to recover watermark picture;

For image file, described locator key frame is: by key frame index, find key frame position from video sequence; For image file, described locator key frame is: directly using image as key frame;

Described watermark fuse information and the decryption information recovery watermark picture of utilizing comprises the following steps: calculate key frame watermark and embed region, extract RGBSan road observational components; Utilize corresponding RGB tri-road separation matrixes to calculate watermarked information source array, and utilize three tunnel information source medium and low frequency vector index to remove to extract three tunnel information source low frequency vectors; The less important dope vector in output San road after utilizing watermark fusion matrix and key frame RGB tri-road low frequency vectors to mix with watermark, recovers watermark information; Wherein, the index of described RGB tri-road separation matrixes and low frequency vector is decryption information, and described fusion matrix and less important dope vector are watermark fuse information.

As fully visible, the present invention proposes the embedding of multimedia digital watermark flexibly and effectively and separation method and device, and user can arrange the feature that watermark embeds according to user's request, comprises the abundant watermark effect such as watermark information, dynamic trajectory and visibility.Watermark directly incorporates in original image or video requency frame data, and strong robustness is difficult for being intercepted even and distorting by illegal means, and unique cipher-text information that after embedding, generation can be resolved watermark is for mutual mandate.

Accompanying drawing explanation

Fig. 1 is the structural representation of multimedia digital watermark processing device in the embodiment of the present invention one;

Fig. 2 is the realization flow figure that in the embodiment of the present invention two, watermark embeds;

Fig. 3 is encoding and decoding streamline schematic diagram in the embodiment of the present invention three;

Fig. 4 is that in the embodiment of the present invention four, digital watermarking is merged and encrypts realization flow figure;

Fig. 5 is that in the embodiment of the present invention five, digital watermarking embeds streamline schematic diagram;

Fig. 6 is the cut-away view schematic diagram of cipher-text information in the embodiment of the present invention six;

Fig. 7 is the realization flow figure of digital watermarking separation in the embodiment of the present invention seven;

Fig. 8 is that in the embodiment of the present invention eight, digital watermark information recovers realization flow figure;

Fig. 9 is the separated streamline schematic diagram of digital watermarking in the embodiment of the present invention nine.

Embodiment

Solution of the present invention is to provide a kind of multimedia digital watermark handling method, and the embedding that can carry out watermark for image, video file is with separated.The method relies on local multimedia solution, image or frame of video are carried out to coding-decoding operation, its processing core is that image or the frame of video to extracting resolved, the mode of using Independent Component Analysis Algorithm to find watermark picture and initial data mutually to merge, and weaken the correlation between video successive frame.Provide multiple watermark to embed form simultaneously, support video dynamic watermark direction, the adjustment of watermark visibility etc.The encryption ciphertext generating is applicable to the reverse process of watermark separation.Corresponding device provides the interface layer making progress, and facilitates local application to use.

The invention provides a kind of multimedia digital watermark embedding method and device, and a kind of multimedia digital watermark separation method and device.

Watermark embedding method is divided into three phases, default stage, watermark processing stage and Dealing with encrypt code stage.

The upper level applications of default stage by the application's device imports watermark embedding demand.Wherein watermark embeds demand, comprises video or image, watermark picture, watermark effect parameter and the ciphertext encrypted ones of local coding and decoding scheme, watermark to be embedded.Watermark effect parameter, comprises the information such as video dynamic watermark, watermark original position, dynamic trajectory, movement rate or visible level.

In the watermark processing stage, comprise startup encoding and decoding streamline, detect key frame, calculate watermark fuse information and decryption information, start watermark embedding streamline, generate the watermarked steps such as media file.

Encoding and decoding streamline wherein, is to open two work sequence of decoding and coding, and decoding work sequence is resolved media file, and reading out data is also exported the frame data of RGB pattern, and the frame data of coding work sequential reception RGB pattern, write in the corresponding media file generating.

Detect key frame, refer to the comparatively uniform frame of video of detected energy distribution in video-frequency band, the object of choosing is to improve watermarking algorithm performance, optimizes watermark and presents effect.If source material is image and non-video is directly processed image as key frame.

Decryption information, comprises key frame RGB tri-road separation matrixes and corresponding three tunnel information source medium and low frequency vector index.Watermark fuse information, comprises that the RGB tri-road low frequency vectors of watermark fusion matrix and key frame mix the less important dope vector in rear output San road with watermark.Calculate watermark fuse information and decryption information, be divided into following steps:

(1) calculate key frame watermark and embed region;

(2) resolve key frame watermark and embed region, extract RGBSan road observational components;

(3) initialization separation matrix;

(4) use FastICA algorithm to calculate corresponding RGB tri-road separation matrixes;

(5) calculate corresponding RGB tri-tunnel information source arrays;

(6) find each road information source medium and low frequency vector index;

(7) according to watermark, embed the requirement of demand to visible level, generate and merge matrix;

(8) watermark information RGB component correspondence be mixed into each road RGB low frequency vector and calculate corresponding RGB tri-road low frequency vectors and the mixed three less important dope vectors in tunnel of watermark.

Start watermark and embed streamline, be divided into following steps:

(1) calculate present frame watermark and embed region;

(2) resolve present frame watermark and embed region, extract RGBSan road observational components;

(3) use RGBSan road separation matrix to calculate RGBSan road information source matrix array;

(4) utilize fusion matrix that information source low frequency vector and watermark vector are merged mutually;

(5) the main information vector after merging is replaced to original low frequency vector, produce the information source after watermarked;

(6) by separated inverse matrix and watermarked after information source calculate the watermarked RGB observational components in three roads.

In the Dealing with encrypt code stage that watermark embeds, comprise that watermark fuse information and decryption information write cryptograph files, use password encryption ciphertext.Ciphertext text comprises watermark effect parameter, watermark key frame index, watermark decryption information and watermark fuse information.

Watermark separation method is also divided into three phases, default stage, Dealing with encrypt code stage and watermark processing stage.

In the default stage, by the upper level applications of the application's device, import local coding and decoding scheme, watermarked image or video, encryption ciphertext, decrypt ciphertext password.

In the Dealing with encrypt code stage, use password decrypting ciphertext, resolve the information in ciphertext.

In the watermark processing stage, comprise startup encoding and decoding streamline, locator key frame, utilize watermark fuse information and decryption information to recover watermark picture, start watermark separated flow waterline and recover original media file.

Locator key frame is from video sequence, to find key frame position by key frame index.If source material is image and non-video is directly processed image as key frame.

Utilize watermark fuse information and decryption information to recover watermark picture, be divided into following steps:

(1) calculate key frame watermark and embed region;

(2) resolve key frame watermark and embed region, extract RGBSan road observational components;

(3) utilize corresponding RGB tri-road separation matrixes to calculate watermarked information source array, and utilize three tunnel information source medium and low frequency vector index to remove to extract three tunnel information source low frequency vectors;

(4) utilize watermark fusion matrix and key frame RGB tri-road low frequency vectors to mix the less important dope vector in rear output San road with watermark, calculate watermark information;

(5) recover watermark picture effect.

Start watermark separated flow waterline, be divided into following steps:

(1) calculate present frame watermark and embed region;

(2) resolve present frame watermark and embed region, extract RGBSan road observational components;

(3) utilize corresponding RGB tri-road separation matrixes to calculate watermarked information source array, and utilize three tunnel information source medium and low frequency vector index to remove to extract three tunnel information source low frequency vectors;

(4) utilize watermark to merge matrix and watermark Vector Message, calculate the three original low frequency vectors in tunnel;

(5) utilize separation matrix inverse matrix to recover the original observation array information in RGBSan road;

In order to make the application's object, technological means and advantage clearer, below in conjunction with accompanying drawing, the application is described in further details.

Embodiment mono-:

If Fig. 1 is the structural representation of multimedia digital watermark processing device in the embodiment of the present invention one.As shown in Figure 1, the device configuring is abstract 101 by user interface, coding and decoding scheme module 102, watermark processing engine 103, encrypt ciphertext module 104 and file storage module 105 forms.In this structure, user interface abstract 101, for docking with application program, receives the parameter that upper strata arranges, and certain application program is described the processing demands of multimedia digital watermark.Coding and decoding scheme module 102 can receive local multimedia scheme, supports the encoding and decoding of video and image, can extract the data in multimedia file, also multi-medium data can be write in the media file of certain form.Watermark processing engine 103 is processing centers of the application's method, also be the core component of the application's device, be used for analyzing the multi-medium data of output from coding and decoding scheme module 102, describe according to demand watermark is embedded in multimedia initial data or from media data separated, for image media, can embed and separated static watermark information, for video media, can also support dynamic watermark, mode with pile line operation is processed, and can support storage and the media data input and output of form online.Encrypting ciphertext module 104 is user interactions authorized component, and the user password according to demand, carries out the information of watermark processing engine 103 outputs textization and encrypt, and for terminal room transmission and authority checking, operates.105 of file storage module are that the file to generating is done stores processor, immediately play or storage, preserve this locality or high in the clouds.

Embodiment bis-:

If Fig. 2 is the realization flow figure that in the embodiment of the present invention two, watermark embeds.From information, import to result output, specifically describe as follows:

Multimedia digital watermark embeds main flow and first enters the default stage, in this stage, by upper level applications, in the application's device, demand parameter is set.

Step 201, imports local coding and decoding scheme demand, makes exactly the application's device get local multimedia processing capability, supports follow-up encoding and decoding work.

Step 202, import video or image resource, this need to squeeze into the original media file of watermark as user, import watermark picture simultaneously, the user interface abstract 101 of the application's device provide receive watermark picture form to upper interface, and process and the look & feel of user's How to choose or establishment watermark pictorial information provide interface to support by upper level applications.

Step 203, imports watermark parameters effect, and this step is exactly the requirement description that heap watermark will be set, and has been provided with watermark picture in upper step 202, will specifically describe the presentation mode of watermark here.

Described watermark presentation mode comprises, is still in the watermark effect on certain fixed position; With certain movement locus mobile dynamic effect etc. that flies.Description comprises the initial position of watermark, dynamical fashion, and dynamic rate, the much informations such as visible level, these information can affect the concrete enforcement of algorithm in follow-up watermark embedding or separation process.

In described step 203, if embedded for the watermark of image resource, do not provide the description of dynamic watermark demand, only provide static mode to present.

Step 204, imports ciphertext password, and this ciphertext password obtains during by upper level applications and user interactions, and object is for being encrypted the process prescription of watermark embedding with password string.While making media file after watermarked to carry out watermark lock out operation on another station terminal, must first through the mandate action between certain user, transmit and encrypt ciphertext, use again password to be decrypted ciphertext, then utilize the ciphertext content after deciphering to carry out watermark separation.Also can receive non-encrypted situation, ciphertext generates with plaintext form.After step 204, carry out the watermark processing stage.

Step 205, before watermark data is processed, starts encoding and decoding streamline.Owing to being provided with before video or the image resource of watermark to be embedded, this step positions resource, and resource may be in this locality also beyond the clouds in server, thus to determine resource acquisition mode, and attempt reading and Study document data.Open two work sequence of decoding and coding, carry out encoding and decoding work to realize the object of real-time conversion and demonstration simultaneously.Specifically refer to tri-descriptions as embodiment.

Step 206, detects key frame.Watermark presentation mode may be static, also may be dynamic, embedding for video watermark, because being directly to revise original media data, namely watermark to be embedded in each frame data, if each frame all calculates separately and produce corresponding decryption information, amount of calculation decryption information huge and that generate is also very huge, is unfavorable for the convenience of real-time and transmission.So the application, in conjunction with FastICA algorithm (Fast-ICA) characteristic, selects key frame to embed benchmark as watermark in sequence of frames of video, and need not be to each frame operation independent all, to reduce amount of calculation and decryption information memory space.

Described key frame, refers to certain frame of video, and its Energy distribution is comparatively even, is exactly overall data intensity profile equilibrium in frame data array, there is no obviously too bright or too dark region effect.Choosing like this, is mainly when implementing Fast-ICA algorithm, data separating successful, and information source frequency distribution also more evenly makes watermark effect better.If source material is image resource, directly image is processed as key frame.

Step 207, calculates key frame watermark fuse information (being called for short watermark fuse information) and decryption information.After finding key frame, by needing watermarked region to resolve to these frame data, find the mode that watermark and initial data are merged mutually, as watermark fuse information, output simultaneously can be used as the data segment of method reverse operation, as decryption information.Specifically refer to following embodiment tetra-descriptions.

Step 208, writes cryptograph files by watermark fuse information and decryption information, these information when separated as watermark must reference one of data be written into cryptograph files.Following embodiment five descriptions of specifically meeting each other.

Step 209, after the content of cryptograph files is determined, using user password as password is encrypted in full to ciphertext, in order to mutual authority of authorizing between safe transmission and user.

Step 210, starts watermark and embeds streamline.Once determine key frame, analyzed watermark fuse information and decryption information, just can carry out watermark embedding to video-frequency band, start the work sequence of streamline form, frame of video is processed one by one.From decoding work sequence, obtain each frame initial data, export watermarked frame data.The following Fig. 6 that specifically meets each other describes.

Consider that user is to watermark demand, may in the process of making watermark video, change watermark picture, need to need to again choose key frame and calculate fuse information and decryption information at corresponding video intersegmental part for the watermark picture upgrading.So described step 206 is to step 210, dotted line frame part, is repetitive process capable of circulation.

Step 211, is to generate watermarked media file, is mainly that the frame data of step 210 output are sent in coding work sequence, writes new media file.

Embodiment tri-:

If Fig. 3 is encoding and decoding streamline schematic diagram in the embodiment of the present invention three, specifically describe as follows:

Decoding work sequence:

Step 301, by local means or long-range means, gets media file data.

Step 302, analyzing medium file format, resolves container and data flow, and the demultiplexing before broadcasting, judges whether this media file can be received and process by local multimedia scheme.

Step 303, if local multimedia scheme cannot be processed this media formats, proceeds to wrong processing, notice upper level applications.

Step 304, if compatible this media formats of local multimedia scheme provides abstract image or video requency frame data.

Step 305, before view data or video requency frame data export watermark processing engine 103 to, is converted into rgb color pattern by its data array, is convenient to later stage algorithm and implements.

Coding work sequence:

Step 306, first coding judges the whether compatible target code form of local multimedia scheme before carrying out.If incompatible, proceed to step 303 and do wrong processing.

Step 307, creates new file destination entity.

Step 308, receives the frame data by the RGB pattern of watermark processing engine 103 outputs.

Step 309, compresses frame data and writes in file destination entity.

Embodiment tetra-:

If Fig. 4 is that in the embodiment of the present invention four, digital watermarking is merged and encrypts realization flow figure, wherein core algorithm is to use Fast-ICA, specifically describes as follows:

Step 401 has got key frame in above-mentioned steps 206, from decoding work sequence, extracts in this step the RGB data of key frame.

Step 402, the watermark demand arranging according to upper layer application, the information such as original position, dynamic trajectory and movement rate, calculate the regional location that watermark embeds this key frame, the region ，Qi area size that namely watermark shows and watermark picture measure-alike.If but watermark picture is mixed with the ratio of viewing area direct to, be easy to utilize inter-frame algorithm to find fusion to stipulate and crack.The application's method is in this viewing area, to carry out implicit expression expansion around, use the mode of extending inquiry that these viewing area data are promoted to hyperspace, the dimension data having more all around extends and from this viewing area, forms multi-dimension array data, is expressed as Observ _key, its dimension is N, the data volume of every one-dimensional space is identical with watermark pictorial information amount, supposes that watermark pictorial information amount is M, Observ _keyfor N * Metzler matrix.Although the increase of dimension can strengthen the robustness of algorithm, also can increase computation complexity simultaneously, in actual algorithm process, generally extend to space-time and can meet algorithm robust performance, do not affect again real-time performance.

Step 403, analyzes Observ _key, therefrom split and extract the array data of corresponding RGB respectively, as the three original observational componentses in tunnel, be expressed as Observ _key-R, Observ _key-Gand Observ _key-B.

Step 404, Fast-ICA algorithm need to be inputted an initial separation matrix before implementing, and it is of a size of N * N, and matrix data obtains at random, is expressed as W _original.

Step 405, after obtaining the initial random separation matrix HeRGB tri-original observational componentses in tunnel, implements Fast-ICA iteration, certain, quantizes to seek in error range solving of isolated component that the degree of correlation is lower, below three formula the relation solving has been described:

Source _key-R＝W _key-R·Observ _key-R (1)

Source _key-G＝W _key-G·Observ _key-G (2)

Source _key-B＝W _key-B·Observ _key-B (3)

According to algorithm principle, under the prerequisite of the observational components of known certain dimension, find out the information source of same dimension, in this information source, between each dimension, mutual statistical is independent.

Step 406, according to the relation that solves in described step 405, i.e. formula (1) (2) (3), algorithm is exported corresponding RGB tri-road separation matrix W _key-R, W _key-G, W _key-B, be N * N square formation.

Step 407 according to the solution formula in described step 405, draws corresponding RGB tri-tunnel information source array Source simultaneously _key-R, Source _key-G, Source _key-B, be N * Metzler matrix.

Step 408, because Fast-ICA is the analytical method based on higher order statistical characteristic, presses frequency band projection by observational components and forms information source component, and information source component medium and low frequency vector the inclusive projection is according to most of energy, respectively at Source _key-R, Source _key-G, Source _key-Bin find the low frequency vector index of its energy maximum, low frequency vector representation is LowFrequency _key-R, LowFrequency _key-G, LowFrequency _key-B, its concordance list is shown Index _key-R, Index _key-G, Index _key-B.

Step 409, the key frame San road separation matrix W of step 406 output _key-R, W _key-G, W _key-Bthe low frequency vector index Index of the corresponding San road information source array of exporting with step 408 _key-R, Index _key-G, Index _key-B, as decryption information, preserve.

Step 410, because the low frequency vector of information source array has comprised most picture region energy, watermark information and this low frequency vector to be carried out to linearity and merge, this linearity merges to be expressed as and merges matrix M ix, due to two one dimension Vector Messages are merged, Mix is two-dimentional square formation.

Described fusion matrix M ix, its value is according to the requirement to watermark visibility in watermark demand.

Step 411, watermark information is expressed as a n dimensional vector n Watermark, is divided into RGBSan road vector Watermark _r, Watermark _g, Watermark _b, following formula has been described fusion process:

$\mathrm{Mix}\·\left[\begin{array}{c}{\mathrm{LowFrequency}}_{\mathrm{key}-R}\\ {\mathrm{Watermark}}_R\end{array}\right]=\left[\begin{array}{c}{\mathrm{Primary}}_{\mathrm{key}-R}\\ {\mathrm{Secondary}}_{\mathrm{key}-R}\end{array}\right]---\left(4\right)$

$\mathrm{Mix}\·\left[\begin{array}{c}{\mathrm{LowFrequency}}_{\mathrm{key}-G}\\ {\mathrm{Watermark}}_G\end{array}\right]=\left[\begin{array}{c}{\mathrm{Primary}}_{\mathrm{key}-G}\\ {\mathrm{Secondary}}_{\mathrm{key}-G}\end{array}\right]---\left(5\right)$

$\mathrm{Mix}\·\left[\begin{array}{c}{\mathrm{LowFrequency}}_{\mathrm{key}-B}\\ {\mathrm{Watermark}}_B\end{array}\right]=\left[\begin{array}{c}{\mathrm{Primary}}_{\mathrm{key}-B}\\ {\mathrm{Secondary}}_{\mathrm{key}-B}\end{array}\right]---\left(6\right)$

Less important dope vector Secondary after output is merged _key-R, Secondary _key-G, Secondary _key-B.

Step 412, less important dope vector Secondary after the fusion of the fusion matrix M ix in step 410 and step 411 output _key-R, Secondary _key-G, Secondary _key-Bas watermark fuse information, preserve.

Embodiment five:

If Fig. 5 is that in the embodiment of the present invention five, digital watermarking embeds streamline schematic diagram, specifically describe as follows:

Step 501 is extracted frame RGB data from decoding work sequence, and note present frame index is n.When if the resource of processing is image, only need to extract frame data.

Step 502, the watermark demand arranging according to upper layer application, the information such as original position, dynamic trajectory and movement rate, calculate the measure-alike of watermark viewing area ，Qi area size and watermark picture.According to the description in described step 402, extend to multi-dimension array data, be expressed as Observ _n, be N * Metzler matrix, N is dimension, M is watermark pictorial information amount.

Step 503, analyzes Observ _n, therefrom split and extract the array data of corresponding RGB respectively, as the three original observational componentses in tunnel, be expressed as Observ _n-R, Observ _n-Gand Observ _n-B.

Step 504, utilizes and has obtained RGB tri-road separation matrix W _key-R, W _key-G, W _key-Bthe ，Dui tri-original observational componentses in tunnel carry out linear operation, corresponding following formula:

Source _n-R＝W _key-R·Observ _n-R (7)

Source _n-G＝W _key-G·Observ _n-G (8)

Source _n-B＝W _key-B·Observ _n-B (9)

Be the equal of in the hyperspace of utilizing key frame to calculate, all use identical projection coefficient to produce corresponding information source component the observational components of each frame, the result output RGBSan road information source array Source of calculating _n-R, Source _n-G, Source _n-B, index and the Index of each information source medium and low frequency component _key-R, Index _key-G, Index _key-Bidentical.

Step 505, is used identical fusion matrix M ix, and watermark is merged to the low frequency vector of information source component, corresponding following formula:

$\mathrm{Mix}\·\left[\begin{array}{c}{\mathrm{LowFrequency}}_{n-R}\\ {\mathrm{Watermark}}_R\end{array}\right]=\left[\begin{array}{c}{\mathrm{Primary}}_{n-R}\\ {\mathrm{Secondary}}_{n-R}\end{array}\right]---\left(10\right)$

$\mathrm{Mix}\·\left[\begin{array}{c}{\mathrm{LowFrequency}}_{n-G}\\ {\mathrm{Watermark}}_G\end{array}\right]=\left[\begin{array}{c}{\mathrm{Primary}}_{n-G}\\ {\mathrm{Secondary}}_{n-G}\end{array}\right]---\left(11\right)$

$\mathrm{Mix}\·\left[\begin{array}{c}{\mathrm{LowFrequency}}_{n-B}\\ {\mathrm{Watermark}}_B\end{array}\right]=\left[\begin{array}{c}{\mathrm{Primary}}_{n-B}\\ {\mathrm{Secondary}}_{n-B}\end{array}\right]---\left(12\right)$

Step 506, watermark is mixed to after information source low frequency vector, and mixed main information vector is replaced to former low frequency vector, forms new information source component arrays Source _n-R', Source _n-G', Source _n-B'.

Step 507, has obtained merging after the information source component after watermark, then it has been converted into new observational components by linear transformation, corresponding following formula:

$W_{\mathrm{key}-R}^{-1}\·{\mathrm{Source}}_{n-R}^{\′}={\mathrm{Observ}}_{n-R}^{\′}---\left(13\right)$

$W_{\mathrm{key}-G}^{-1}\·{\mathrm{Source}}_{n-G}^{\′}={\mathrm{Observ}}_{n-G}^{\′}---\left(14\right)$

$W_{\mathrm{key}-B}^{-1}\·{\mathrm{Source}}_{n-B}^{\′}={\mathrm{Observ}}_{n-B}^{\′}---\left(15\right)$

Step 508, generates new observational components array Observ _n-R', Observ _n-G', Observ _n-B' as the frame data region after watermarked, the original area data before covering, thus form the frame data after new watermarked of a frame.

Step 509, by generate watermarked after frame data send into coding work sequence.

Embodiment six:

If Fig. 6 is the cut-away view schematic diagram of cipher-text information in the embodiment of the present invention six, comprise watermark effect parameter 601, watermark key frame index 602, watermark decryption information 603 and watermark fuse information 604.

Described watermark effect parameter 601, is the watermark demand that upper layer application arranges, and comprises the information such as video dynamic watermark, watermark original position, dynamic trajectory, movement rate or visible level.

Described watermark key frame index 602 is key frame index positions in sequence of frames of video.

Described watermark decryption information 603, comprises step 406 output San road separation matrix W _key-R, W _key-G, W _key-Bthe low frequency vector index Index of the corresponding three tunnel information source arrays of exporting with step 408 _key-R, Index _key-G, Index _key-B.

Described watermark fuse information 604, less important dope vector Secondary after the fusion that the fusion matrix M ix that comprises in step 410 output and step 411 are exported _key-R, Secondary _key-G, Secondary _key-B.

Wherein said watermark key frame index 602, watermark decryption information 603 and watermark fuse information 604, can, according to the quantity that watermark picture in demand is set, increase and decrease accordingly.

Embodiment seven:

If Fig. 7 is the realization flow figure of digital watermarking separation in the embodiment of the present invention seven, from information, import to result output equally, specifically describe as follows:

Step 701, imports local coding and decoding scheme demand, makes the application's device get local multimedia processing capability.

Step 702, imports watermarked video or image resource.

Step 703, imports the encryption ciphertext generated by watermark embed process, and how this ciphertext is transmitted or how to be shared at terminal room and do not belong to the application's coverage.

Step 704, imports user password, if ciphertext text is encrypted via user password in the process generating, when resolving, needs corresponding password as deciphering voucher, and how user password does not belong to the application's coverage in terminal room transmission.

Step 705, is used password to be decrypted processing to encrypting ciphertext, obtains the clear content of cryptograph files.

Step 706, resolves ciphertext, from content, analyzes and obtains described watermark effect parameter 601, watermark key frame index 602, watermark decryption information 603 and watermark fuse information 604.

Step 707, proceed to the watermark processing stage, first start encoding and decoding streamline, owing to being provided with video or the image resource of watermark to be separated, this step positions resource, resource may be in this locality also beyond the clouds in server, thus to determine resource acquisition mode, and attempt reading and Study document data.Open two work sequence of decoding and coding, carry out encoding and decoding work to realize real-time conversion and demonstration simultaneously.

Step 708, the watermark key frame index 602 according to having obtained is directly targeted to key frame position in video sequence.When the media resource of watermark to be separated, during image, key frame is exactly this image itself.

Step 709, according to the watermark effect parameter 601 of having obtained, watermark decryption information 603 and watermark fuse information 604, resolves key frame.Extract watermark region, and the use inverse process recovery watermark pictorial information contrary with telescopiny, specifically describe the following Fig. 8 that meets each other and describe.

Step 710, starts watermark and embeds streamline.Use watermark fuse information and decryption information, video-frequency band is carried out to watermark separation, start the work sequence of streamline form, frame of video is processed one by one.From decoding work sequence, obtain the watermarked data of each frame, the original frame data after output watermark.Specifically refer to following embodiment nine descriptions.

The change possibility of considering watermark picture in video sequence, described step 708 is to step 710, and dotted line frame part, is repetitive process capable of circulation.

Step 711, is to recover original media file, is mainly that the frame data of step 710 output are sent in coding work sequence, writes new media file.

Embodiment eight:

If Fig. 8 is that in the embodiment of the present invention eight, digital watermark information recovers realization flow figure, specifically describe as follows:

Step 801, above-mentioned steps 708 has been obtained key frame index position, extracts the RGB data of key frame from decoding work sequence.

Step 802, in watermark demand, the information such as original position, dynamic trajectory and movement rate, calculates the position, viewing area that watermark embeds this key frame.And by this viewing area data extending to hyperspace, the dimension data having more around extends and from this viewing area according to certain rule, forms multi-dimension array data, this array is exactly corresponding to Observ described in step 402 _keywatermarked array, be expressed as Observ _key', be N * Metzler matrix.

Step 803, by Observ _key' be split as watermarked RGB tri-road observational components Observ _key-R', Observ _key-G', Observ _key-B'.

Step 804, goes for the low frequency vector of information source component according to the watermark decryption information 603 extracting from ciphertext, namely watermarked that vector at that time.First by RGBSan road observational components Observ _key-R', Observ _key-G', Observ _key-B' HeRGB tri-road separation matrix W _key-R, W _key-G, W _key-B, by linear combination, calculating information source array, corresponding formula is as follows:

Source _key-R′＝W _key-R·Observ _key-R′ (16)

Source _key-G′＝W _key-G·Observ _key-G′ (17)

Source _key-B′＝W _key-B·Observ _key-B′ (18)

Again by the low frequency vector index Index of RGBSan road information source array _key-R, Index _key-G, Index _key-B, find corresponding information source array RGB tri-road low frequency vector Primary _key-R, Primary _key-G, Primary _key-B.

Step 805, removes to recover watermark information picture according to the watermark fuse information 604 extracting from ciphertext.Namely use and merge matrix M ix and merge rear less important dope vector Secondary _key-R, Secondary _key-G, Secondary _key-B, utilize above-mentioned formula (4) (5) (6) to calculate watermark information RGBSan road vector Watermark _r, Watermark _g, Watermark _b, can merge into watermark pictorial information vector Watermark.

Step 806, is converted into watermark picture by watermark information vector Watermark, can be for showing or secondary editor.

Embodiment nine:

If Fig. 9 is the separated streamline schematic diagram of digital watermarking in the embodiment of the present invention nine, specifically describe as follows:

Step 901 is extracted frame RGB data from decoding work sequence, and note present frame index is n.When if the resource of processing is image, only need to extract frame data.

Step 902, according to the watermark effect parameter 601 of extracting from ciphertext, comprise the information such as video dynamic watermark, watermark original position, dynamic trajectory, movement rate or visible level, calculate watermark viewing area, according to the description in described step 402, extend to multi-dimension array data, be expressed as Observ _n', be N * Metzler matrix, N is dimension, M is watermark pictorial information amount.

Step 903, analyzes Observ _n', therefrom split and extract the array data of corresponding RGB respectively, as the watermarked observational components in three roads, be expressed as Observ _n-R', Observ _n-G' and Observ _n-B'.

Step 904, for the observational components of present frame, goes for the low frequency vector of information source component according to the watermark decryption information 603 extracting from ciphertext, namely watermarked data vector at that time.First utilize RGBSan road observational components Observ _n-R', Observ _n-G' and Observ _n-B' HeRGB tri-road separation matrix W _key-R, W _key-G, W _key-B, by linear combination, calculating information source array, corresponding formula is as follows:

Source _n-R′＝W _key-R·Observ _n-R′ (19)

Source _n-G′＝W _key-G·Observ _n-G′ (20)

Source _n-B′＝W _key-B·Observ _n-B′ (21)

Again by the low frequency vector index Index of RGBSan road information source array _key-R, Index _key-G, Index _key-B, find the corresponding information source array of present frame observational components RGB tri-road low frequency vector Primary _n-R, Primary _n-G, Primary _n-B.

Step 905, according to described step 805, has extracted watermark information vector Watermark _r, Watermark _g, Watermark _b, the fusion matrix M ix in known ciphertext fuse information, adds information source array RGB tri-road low frequency vector Primary simultaneously _n-R, Primary _n-G, Primary _n-B, can utilize above-mentioned formula (10) (11) (12) to calculate original low frequency vector LowFrequency when not watermarked _n-R, LowFrequency _n-G, LowFrequency _n-B

Step 906, in the information source of present frame observational components, with original low frequency vector LowFrequency _n-R, LowFrequency _n-G, LowFrequency _n-Bremove to replace Primary _n-R, Primary _n-G, Primary _n-B, obtain original source array data Source when not watermarked _n-R, Source _n-G, Source _n-B.

Utilize formula (7) (8) (9) to calculate original observation array Observ when not watermarked _n-R, Observ _n-Gand Observ _n-B.

Step 907 ，RGB tri-tunnel original observation array Observ _n-R, Observ _n-Gand Observ _n-Bbe merged into Observ _nthereby, recover original image or video requency frame data, supply to show or send into coding work sequence.

The present invention also proposes a kind of multimedia digital watermark flush mounting, comprising:

User interface is abstract, for importing watermark, embeds demand; Described watermark embedding demand comprises watermark effect parameter; Described watermark effect parameter comprises video dynamic watermark, watermark original position, dynamic trajectory, movement rate or visible level;

Watermark processing engine, for calculating watermark fuse information and decryption information; The step of described calculating watermark fuse information and decryption information comprises: calculate key frame watermark and embed region, extract RGBSan road observational components; Use FastICA algorithm to calculate corresponding RGB tri-road separation matrixes; Calculate corresponding RGB tri-tunnel information source arrays; Find the low frequency vector index in each road information source matrix; According to watermark, embed the requirement of demand to visible level, generate and merge matrix; Watermark information RGB component correspondence is mixed into each road RGB low frequency vector of key frame and calculates corresponding RGB tri-road low frequency vectors and the mixed three less important dope vectors in tunnel of watermark; Described RGB tri-road separation matrixes and low frequency vector index are decryption information, and described fusion matrix and less important dope vector are watermark fuse information;

Encrypt ciphertext module, for carrying out Dealing with encrypt code, the ciphertext of described Dealing with encrypt code comprises: watermark effect parameter, watermark key frame index, decryption information and watermark fuse information.

Said apparatus can also comprise:

Coding/decoding module, for opening decoding or coding work sequence;

Described decoding work sequence is: resolve media file, reading out data is also exported the frame data of RGB pattern;

Described coding work sequence is: receive the frame data of RGB pattern, write in the corresponding media file generating.

The present invention also proposes a kind of multimedia digital watermark separator, comprising:

User interface is abstract, default for carrying out;

Encrypt ciphertext module, for carrying out Dealing with encrypt code;

Watermark processing engine, for carrying out watermark processing, comprising: locator key frame, utilize watermark fuse information and decryption information to recover watermark picture;

For image file, described locator key frame is: by key frame index, find key frame position from video sequence; For image file, described locator key frame is: directly using image as key frame;

Described watermark fuse information and the decryption information recovery watermark picture of utilizing comprises the following steps: calculate key frame watermark and embed region, extract RGBSan road observational components; Utilize corresponding RGB tri-road separation matrixes to calculate watermarked information source array, and utilize three tunnel information source medium and low frequency vector index to remove to extract three tunnel information source low frequency vectors; The less important dope vector in output San road after utilizing watermark fusion matrix and key frame RGB tri-road low frequency vectors to mix with watermark, recovers watermark information; Wherein, the index of described RGB tri-road separation matrixes and low frequency vector is decryption information, and described fusion matrix and less important dope vector are watermark fuse information.

In the method that the present invention proposes, watermark embedding and separation process amount of calculation are little real-time, can adapt to online streaming media playing.The device that the present invention proposes can be deployed to various terminals, can be used by various application program, is convenient to realize embedding, the separation of multimedia digital watermark and the end-to-end interactive mode of editing again.

The foregoing is only preferred embodiment of the present invention, in order to limit the present invention, within the spirit and principles in the present invention not all, any modification of making, be equal to replacement, improvement etc., within all should being included in the scope of protection of the invention.

Claims (15)

1. a multimedia digital watermark embedding method, is characterized in that, described method comprises: default stage, watermark processing stage and Dealing with encrypt code stage;

The described default stage comprises: import watermark and embed demand; Described watermark embedding demand comprises watermark effect parameter; Described watermark effect parameter comprises video dynamic watermark, watermark original position, dynamic trajectory, movement rate or visible level;

The described watermark processing stage comprises the step of calculating watermark fuse information and decryption information; The step of described calculating watermark fuse information and decryption information comprises: calculate key frame watermark and embed region, extract RGBSan road observational components; Use FastICA algorithm to calculate corresponding RGB tri-road separation matrixes; Calculate corresponding RGB tri-tunnel information source arrays; Find the low frequency vector index in each road information source matrix; According to watermark, embed the requirement of demand to visible level, generate and merge matrix; Watermark information RGB component correspondence is mixed into each road RGB low frequency vector of key frame and calculates corresponding RGB tri-road low frequency vectors and the mixed three less important dope vectors in tunnel of watermark; Described RGB tri-road separation matrixes and low frequency vector index are decryption information, and described fusion matrix and less important dope vector are watermark fuse information;

The ciphertext in described Dealing with encrypt code stage comprises: watermark effect parameter, watermark key frame index, decryption information and watermark fuse information.

2. method according to claim 1, is characterized in that, described watermark embeds demand and also comprises: the video of local coding and decoding scheme, watermark to be embedded or image, watermark picture and ciphertext encrypted ones.

3. method according to claim 1, is characterized in that, described watermark processing is in the stage, and the step of calculating watermark fuse information and decryption information also comprises before: start encoding and decoding streamline, detect key frame;

After the step of described calculating watermark fuse information and decryption information, also comprise: start watermark and embed streamline, generate watermarked media file.

4. method according to claim 3, is characterized in that, described encoding and decoding streamline is for opening decoding or coding work sequence;

Described decoding work sequence is: resolve media file, reading out data is also exported the frame data of RGB pattern;

Described coding work sequence is: receive the frame data of RGB pattern, write in the corresponding media file generating.

5. method according to claim 3, is characterized in that, for video file, described detection key frame is: the comparatively uniform frame of video of detected energy distribution in video-frequency band, using this frame of video as key frame;

For image file, described detection key frame is: directly using described image as key frame.

6. method according to claim 3, is characterized in that, described startup watermark embeds streamline and comprises the following steps:

(1) calculate present frame watermark and embed region;

(2) resolve present frame watermark and embed region, extract RGBSan road observational components;

(3) use RGBSan road separation matrix to calculate RGBSan road information source matrix array;

(4) utilize fusion matrix that information source low frequency vector and watermark vector are merged mutually;

(5) the main information vector after merging is replaced to original low frequency vector, produce the information source after watermarked;

(6) by separated inverse matrix and watermarked after information source calculate the watermarked RGB observational components in three roads.

7. method according to claim 1, is characterized in that, the described Dealing with encrypt code stage comprises: watermark fuse information and decryption information are write to cryptograph files and use password encryption ciphertext.

8. a multimedia digital watermark separation method, is characterized in that, described method comprises: default stage, Dealing with encrypt code stage and watermark processing stage;

The described watermark processing stage comprises: locator key frame, utilize watermark fuse information and decryption information to recover watermark picture;

For image file, described locator key frame is: by key frame index, find key frame position from video sequence; For image file, described locator key frame is: directly using image as key frame;

Described watermark fuse information and the decryption information recovery watermark picture of utilizing comprises the following steps: calculate key frame watermark and embed region, extract RGBSan road observational components; Utilize corresponding RGB tri-road separation matrixes to calculate watermarked information source array, and utilize three tunnel information source medium and low frequency vector index to remove to extract three tunnel information source low frequency vectors; The less important dope vector in output San road after utilizing watermark fusion matrix and key frame RGB tri-road low frequency vectors to mix with watermark, recovers watermark information; Wherein, the index of described RGB tri-road separation matrixes and low frequency vector is decryption information, and described fusion matrix and less important dope vector are watermark fuse information.

9. method according to claim 8, is characterized in that, the described default stage comprises: import local coding and decoding scheme, watermarked image or video, encryption ciphertext, decrypt ciphertext password.

10. method according to claim 8, is characterized in that, the described Dealing with encrypt code stage comprises the information of using in password decrypting ciphertext, parsing ciphertext.

11. methods according to claim 8, is characterized in that, the described watermark processing stage also comprises: start encoding and decoding streamline, start watermark separated flow waterline and recover original media file.

12. methods according to claim 11, is characterized in that, described startup watermark separated flow waterline comprises the following steps:

(1) calculate present frame watermark and embed region;

(2) resolve present frame watermark and embed region, extract RGBSan road observational components;

(3) utilize corresponding RGB tri-road separation matrixes to calculate watermarked information source array, and utilize three tunnel information source medium and low frequency vector index to remove to extract three tunnel information source low frequency vectors;

(4) utilize watermark to merge matrix and watermark Vector Message, calculate the three original low frequency vectors in tunnel;

(5) utilize separation matrix inverse matrix to recover the original observation array information in RGBSan road.

13. 1 kinds of multimedia digital watermark flush mountings, is characterized in that, described device comprises:

User interface is abstract, for importing watermark, embeds demand; Described watermark embedding demand comprises watermark effect parameter; Described watermark effect parameter comprises video dynamic watermark, watermark original position, dynamic trajectory, movement rate or visible level;

Watermark processing engine, for calculating watermark fuse information and decryption information; The step of described calculating watermark fuse information and decryption information comprises: calculate key frame watermark and embed region, extract RGBSan road observational components; Use FastICA algorithm to calculate corresponding RGB tri-road separation matrixes; Calculate corresponding RGB tri-tunnel information source arrays; Find the low frequency vector index in each road information source matrix; According to watermark, embed the requirement of demand to visible level, generate and merge matrix; Watermark information RGB component correspondence is mixed into each road RGB low frequency vector of key frame and calculates corresponding RGB tri-road low frequency vectors and the mixed three less important dope vectors in tunnel of watermark; Described RGB tri-road separation matrixes and low frequency vector index are decryption information, and described fusion matrix and less important dope vector are watermark fuse information;

Encrypt ciphertext module, for carrying out Dealing with encrypt code, the ciphertext of described Dealing with encrypt code comprises: watermark effect parameter, watermark key frame index, decryption information and watermark fuse information.

14. devices according to claim 13, is characterized in that, described device also comprises:

Coding/decoding module, for opening decoding or coding work sequence;

Described decoding work sequence is: resolve media file, reading out data is also exported the frame data of RGB pattern;

Described coding work sequence is: receive the frame data of RGB pattern, write in the corresponding media file generating.

15. 1 kinds of multimedia digital watermark separators, is characterized in that, described device comprises:

User interface is abstract, default for carrying out;

Encrypt ciphertext module, for carrying out Dealing with encrypt code;

Watermark processing engine, for carrying out watermark processing, comprising: locator key frame, utilize watermark fuse information and decryption information to recover watermark picture;

For image file, described locator key frame is: by key frame index, find key frame position from video sequence; For image file, described locator key frame is: directly using image as key frame;

Described watermark fuse information and the decryption information recovery watermark picture of utilizing comprises the following steps: calculate key frame watermark and embed region, extract RGBSan road observational components; Utilize corresponding RGB tri-road separation matrixes to calculate watermarked information source array, and utilize three tunnel information source medium and low frequency vector index to remove to extract three tunnel information source low frequency vectors; The less important dope vector in output San road after utilizing watermark fusion matrix and key frame RGB tri-road low frequency vectors to mix with watermark, recovers watermark information; Wherein, the index of described RGB tri-road separation matrixes and low frequency vector is decryption information, and described fusion matrix and less important dope vector are watermark fuse information.

Images