TW201413637A - Method for embedding and extracting digital watermarks - Google Patents

Abstract

Method for embedding digital watermarks includes: inputting a first image then transforming the first images' pixel from a preset color mode to YUV mode, calculating an illumination function dividing the first images' pixel to a plurality of planes equally, calculating an level function dividing the plurality of plane's pixel to a plurality of groups, calculating an hue function dividing the plurality of group's pixel to a plurality of blocks, hiding the digital watermarks into the plurality of blocks corresponsively, transforming the first images' pixel from YUV mode to the preset color mode, outputting the first image hided the digital watermarks to be a second image.

Description

Method of embedding and extracting digital watermark

The present invention relates to a watermarking technique, and more particularly to a method of embedding and extracting a digital watermark.

Digital watermarking is a way to protect multimedia content from unauthorized playback and reproduction. Digital watermarking technology refers to adding a copyright representative's logo (such as a registered trademark, a for-profit business number, a personal portrait, etc.) to the protected digital information to represent and prove the legal possession of the digital multimedia information. The attribution of rights. Moreover, after the digital information embedded in the digital watermark is processed and destroyed by some common digital signals, and the quality is still within an acceptable range, the digital watermark can still be displayed, so that the digital information can be proved. Copyright is vested to detect misappropriation or counterfeiting by illegal users.

In 2000, Lu et al. proposed a watermarking technique based on vector quantization (VQ) for compressed embedding mode. In this method, before compressing the image and implanting the watermark, the VQ codebook must be appropriately expanded to generate more codewords and the codewords of similar content will belong to the same group.

When a watermark bit is to be embedded in an image block, the implanted action is performed based on the position of the block's compression vector within its group. In this way, this method achieves both compression and protection. However, the above expanded code book must It must be kept strictly confidential and cannot be made public, otherwise its security will be threatened. Subsequently, Makur and Selvi scholars proposed another watermarking technique based on vector quantization coding, which uses a variable dimension vector quantization (VDVQ).

That is to say, a code book will contain code words of two dimensions, and will be divided into two groups according to different dimensions. Then, according to the watermark bit to be embedded, it is determined from which group to select the most similar code word for the compression operation. However, this method has a drawback in that the image embedded in the watermark is more fragile, and the unscrupulous person is easy to find the location of the watermark, and then removes and reproduces.

Therefore, how to design a simple and effective invisible watermarking technology based on the development of compression code, while the embedded image still has confidentiality and aesthetics, which has become the focus of its research.

The existing digital watermarking technology can be divided into two types according to the way of watermark hiding. The first type is to hide the digital watermark in the frequency domain of the image, such as the region after the image is transformed by discrete cosine transform or wavelet. However, this method has a high calculation amount, so the execution speed is low. The second type is to hide the digital watermark in the spatial domain of the image, such as: hiding the watermark directly in the RGB or YUV pixel values of the 2D image. However, most of these methods are weak against geometric attacks. For example, after the image is cropped, the watermark information may be removed.

A new digital watermarking algorithm is proposed to guarantee the intellectual property rights of digital image data.

The method of embedding the digital watermark includes inputting the first image, converting the image pixel of the first image into a YUV color model from the preset color model, and calculating the brightness function to divide the image pixels of the first image into a plurality of planes, and calculating the concentration The function divides the image pixels of each plane into a plurality of groups, and the calculated chromaticity function divides the image pixels of each group into a plurality of blocks, and the watermark data is correspondingly hidden in the block. The first image is converted from the YUV color model to a preset color model, and the first image in which the watermark data is stored is output as the second image.

The method for retrieving the digital watermark includes inputting the first image, converting the image pixels of the first image into a YUV color model from a preset color model, and calculating a luminance function to divide the image pixels of the first image into a plurality of planes, and calculating The density function divides the image pixels of each plane into a plurality of groups, and the calculated chromaticity function divides the image pixels of each group into a plurality of blocks, and calculates the feature function to each of the luminance functions. The plane is the basis of calculation, and the root mean square value of the gradient values of the adjacent plurality of directions of each image pixel of the luminance plane is calculated, and the binary bit value of the watermark contained in each block is extracted, and the comparison is performed. The binary bit value of the extracted watermark and the watermark data, wherein when the binary bit value of the extracted watermark reaches the critical value of the watermark data, it indicates that the watermark data exists in the first image.

The invention inserts the digital watermark data by using the conversion format, and does not cause too much change to the original image, and is not easy to be perceived by the naked eye, and can resist geometric attacks such as image enlargement, reduction, rotation, cropping, various image compression. (eg JPEG), sharpness processing, sharpness processing, changing aspect ratio, reducing color, noise filters, etc. or It is a comprehensive attack, and can effectively restore the hidden digital watermark data to protect the intellectual property rights of digital image data.

The detailed features and advantages of the present invention are set forth in the Detailed Description of the Detailed Description of the <RTIgt; </ RTI> <RTIgt; </ RTI> </ RTI> </ RTI> <RTIgt; The objects and advantages associated with the present invention can be readily understood by those skilled in the art.

1 to 2 are flow charts of the embedded digital watermark of the present invention. In an embodiment of the present invention, the digital watermark data W(t) is embedded in the first image, and the length of the first image is set to M, the width is set to N, and the digital watermark data to be embedded is W(t). ).

Wherein, the digital watermark data W(t) can be converted into a digital watermark data W(t) represented by a binary value by a watermark of a color or gray scale. The digital watermark data W(t) is a binary value of length T. Wherein, the length T represents the bit length of the digital watermark data, and t represents the value of the t-th bit of the digital watermark data W(t).

For example, if W(t)=0100001011...1, 0<=t<T, T=100, it can be seen that the bit length T of the watermark data is 100 bits, and the first bit of the watermark data W (0) = 0, the second bit W (1) = 1, ... is the same as the 100th bit W (99) = 1.

The method of embedding a digital watermark includes inputting a first image (S10), and converting image pixels of the first image into a YUV color model from a preset color model (S11). The preset color model may be an RGB color model.

The RGB color models are represented by red, green, and blue, respectively. Plane Y is an image The luminance signal in the plane, and the plane U is the chromaticity in the image, and the plane V is the density in the image. Among them, U and V are also called chroma signals.

Here, the luminance function Y _f (x, y) is calculated to divide the image pixels of the first image into a plurality of planes (S12). Wherein, the luminance function Y _f (x, y) is calculated based on the luminance plane (Y plane) in the YUV color model, and the calculation is performed with each image pixel Y(x, y) of the luminance plane as a center, and Set the average of all image pixels within radius R.

Therefore, the luminance function Y _f (x, y) has an area of the predetermined radius R of N _R , and the average value of all image pixels in the area of N _R can be expressed as: 0<=x<=M-1 , and 0<=y<=N-1, and i and j are genus and integer, and i ² + j ² < R ² , then Y _f (x, y)=

In this embodiment, a plurality of planes are p planes. Where p is an integer. Here, all the pixels of the first image are multiplied by a wide number of pixels, that is, M ^* N pixels. Therefore, each plane is (M ^* N) / p pixels.

And, the calculated density function V _f (x, y) divides the image pixels of each plane into a plurality of groups (S13). Wherein, the concentration function V _f (x, y) is based on the concentration plane (V plane) in the YUV color model, and the calculation is based on the image plane U(x, y) of the concentration plane as the center, and Set the average of all image pixels within radius R.

Therefore, the area of the concentration function V _f (x, y) at the preset radius R is N _R , and the average value of all image pixels in the area N _R can be expressed as: 0<=x<=M-1 , and 0<=y<=N-1, and i and j are genus and integer, and i ² + j ² < R ² , then V _f (x, y)=

Here, the plurality of groups are q groups. Where q is an integer. Therefore, each group is (M ^* N) / (p ^* q) pixels.

And, the calculated chromaticity function U(x, y) divides the image pixels of each group into a plurality of blocks (S14). Among them, the chromaticity function U(x, y) is the chromaticity plane (U plane) in the YUV color model.

Here, a plurality of blocks are set as r blocks. Where r is an integer.

And, the feature function CG(x, y) is calculated, and the feature function CG(x, y) is calculated based on each plane of the luminance function Y _f (x, y), and each image of the luminance plane is calculated. The root mean square value after the gradient value of the plurality of directions adjacent to the pixel Y (x, y) (S15).

Please refer to Figure 3, for example, when the adjacent direction can be right ( g _h ), lower right ( g _d1 ), lower ( g _v ), and lower left ( g _d2 ), the characteristic function CG(x, y) can be expressed as: .

Based on this, the luminance function Y _f (x, y), the density function V _f (x, y), the chromaticity function U (x, y), and the eigenfunction of each pixel in the first image can be calculated. CG(x, y). And, all the pixels in the first image are divided into p ^* q ^* r blocks, and each block is (M ^* N) / (p ^* q ^* r) pixels.

Here, each block divides the chrominance value (U value) of each block into four by the distribution of image pixels in the chromaticity function U(x, y) and the feature function CG(x, y). The chrominance region adjusts the chrominance value of the image pixel of the first image, and represents the binary bit value of the watermark data hidden in the block by the chromaticity value of the image pixel of the adjusted first image. However, the present invention is not limited thereto, and the number of chrominance regions can be adjusted as needed.

At this time, the watermark data is correspondingly hidden in the block (S16). That is to say, the digital watermark data W(t) of one bit is separately hidden in one block. Here, the image pixels of each block are based on the luminance function Y _f (x, y), the density function V _f (x, y), the chromaticity function U (x, y), and the feature function CG ( x, y) are distinguished by four feature functions.

For example, if the image pixel of the first image is divided into 100 blocks, the digital watermark data W(t) of one bit is hidden in one block to adjust each block of the first image. The image pixels are such that the changed block contains digital watermark data W(t).

In addition, if the image pixel of the first image is divided into 1000 blocks, and the digital watermark data W(t) is a binary value of 100 bits, the corresponding watermark data W(t) is calculated. The block of the first image, and the digital watermark data W(t) of one bit is hidden in a block.

In the process of adjusting the chrominance value (U value) of each block in the image pixel in the block of the first image, the block is based on the chromaticity function U(x, y) and the feature function CG (x, y) cut into four chrominance regions N _A , N _B , N _C , N _D , such as the chromaticity function U(x, y) is the x-axis, and the characteristic function CG(x, y) is the y-axis. And the binary value data of the t-th bit in the digital watermark data W(t) is hidden in the block, so that the block represents the value of the digital watermark data W(t) according to the image pixel distribution. .

Here, 0 < α < 1 is expressed as the intensity of the watermark. If W(t)=1, some pixels are moved from the blocks N _B and N _C into the blocks N _A and N _D such that N _A , N _D >=(1+α) and N _B , N _C <=( 1-α). If W(t) = 0, some pixels are moved from the blocks N _A , N _D into the blocks N _B , N _C such that N _B , N _C >=(1+α) and N _A , N _D <=( 1-α).

Thereby, the adjusted image pixels are in the luminance function Y _f (x, y), the density function V _f (x, y), the chromaticity function U (x, y), and the feature function CG (x The distribution pattern in , y) represents the binary value of the digital watermark data W(t).

Here, the first image is converted from the YUV color model to a preset color model (S17), and the first image in which the watermark data is stored is output as the second image (S18).

4 to 5 are flow charts of the digital watermarking of the present invention.

The method for retrieving the digital watermark includes inputting the first image (S20), converting the image pixels of the first image into an YUV color model (S21), and calculating the luminance function to divide the image pixels of the first image into equal parts. A plurality of planes (S22), the calculation density function divides the image pixels of each plane into a plurality of groups (S23), and the calculated chromaticity function divides the image pixels of each group into a plurality of blocks ( S24), the calculation feature function calculates a root mean square value of a gradient value of a plurality of adjacent directions of each image pixel of the luminance plane based on each plane of the luminance function (S25).

Here, the detailed description of the method of retrieving the digital watermark is the same as embedding the digital floating Watermark.

Thereafter, the binary bit value of the watermark contained in each block is extracted (S26), and the binary bit value of the extracted watermark is compared with the watermark data (S27), wherein the extracted watermark binary When the bit value reaches the critical value of the watermark data (S28), it means that the watermark data exists in the first image (S29). Otherwise, it means that there is no watermark data in the first image (S30).

The threshold value represents the binary bit value of the extracted watermark in accordance with the ratio of the watermark data, such as more than 60 percent.

The invention inserts the digital watermark data by using the conversion format, and does not cause too much change to the original image, and is not easy to be perceived by the naked eye, and can resist geometric attacks such as image enlargement, reduction, rotation, cropping, various image compression. (such as JPEG), sharpness processing, sharpness processing, changing the aspect ratio, reducing color, noise filter, etc. or comprehensive attacks, and can effectively restore the hidden digital watermark data to protect digital image data. Intellectual property rights.

Although the technical content of the present invention has been disclosed in the above preferred embodiments, it is not intended to limit the present invention, and any modifications and refinements made by those skilled in the art without departing from the spirit of the present invention are encompassed by the present invention. The scope of protection of the present invention is therefore defined by the scope of the appended claims.

FIG. 1 is a flow chart (1) of embedding a digital watermark according to an embodiment of the present invention.

2 is a flow chart (2) of embedding a digital watermark according to an embodiment of the present invention.

Figure 3 is a schematic diagram of a characteristic function of an embodiment of the present invention.

FIG. 4 is a flow chart (1) of taking out a digital watermark according to an embodiment of the present invention.

FIG. 5 is a flow chart (2) of taking out a digital watermark according to an embodiment of the present invention.

Claims (13)

A method for embedding a digital watermark comprises: inputting a first image; converting an image pixel of the first image into a YUV color model by using a preset color model; calculating a luminance function to image pixels of the first image Dividing into a plurality of planes; calculating a concentration function to divide each image pixel of the plane into a plurality of groups; calculating a chrominance function to divide each image group of the group into a plurality of blocks; A watermark data is correspondingly hidden in the blocks; the first image is converted into the preset color model by the YUV color model; and the first image in which the watermark data is stored is a second image. The method for embedding a digital watermark as described in claim 1, wherein the calculating the chromaticity function further comprises: calculating a feature function based on the planes, and calculating a phase of each image pixel of the brightness plane. The root mean square value of the gradient value of the neighboring multiple directions. The method of embedding a digital watermark as described in claim 2, wherein the directions are right, bottom right, bottom, and bottom left. The method for embedding a digital watermark as described in claim 2, wherein the watermark data is correspondingly hidden in the blocks, comprising: adjusting image pixels of the first image After the chrominance value, each of the blocks represents the binary bit value of the watermark data hidden by the block in the chromaticity function and the distribution pattern of the image pixels in the feature function. The method for embedding a digital watermark as described in claim 1, wherein the luminance function is based on a luminance plane in a YUV color model, and is calculated as a center of each image pixel of the luminance plane, and the preset radius is The average of all image pixels within R. The method for embedding a digital watermark as described in claim 1, wherein the concentration function is based on a concentration plane in the YUV color model, and the pixel is calculated as a center of each image pixel of the density plane, and the preset radius is The average of all image pixels within R. A method of embedding a digital watermark as described in claim 1, wherein the chromaticity function is a chromaticity plane in a YUV color model. The method of embedding a digital watermark as described in claim 1, wherein the preset color model is an RGB color model. A method for retrieving a digital watermark comprises: inputting a first image; converting an image pixel of the first image from an RGB color model to a YUV color model; calculating a luminance function to equally divide the image pixels of the first image a plurality of planes; calculating a density function to divide the image pixels of each plane into a plurality of a group; calculating a chrominance function to divide each image group of the group into a plurality of blocks; calculating a feature function based on the planes, and calculating the neighboring of each image pixel of the brightness plane The root mean square value of the gradient values of the plurality of directions; extracting the binary bit values of the watermarks contained in each of the blocks; comparing the binary bit values of the extracted watermarks with a watermark data, When the binary bit value of the extracted watermark reaches a critical value corresponding to the watermark data, it indicates that the watermark data exists in the first image. The method of embedding a digital watermark as described in claim 9, wherein the directions are right, bottom right, bottom, and bottom left. The method of embedding a digital watermark as described in claim 9, wherein the luminance function is calculated based on a luminance plane in the YUV color model, and the calculation is performed by using each image pixel of the luminance plane as a center, and the preset radius is The average of all image pixels within R. The method of embedding a digital watermark as described in claim 9, wherein the concentration function is based on a concentration plane in the YUV color model, and the calculation is performed by using each image pixel of the density plane as a center, and the preset radius is The average of all image pixels within R. A method of embedding a digital watermark as described in claim 9, wherein the chromaticity function is a chromaticity plane in a YUV color model.