KR100369306B1 - Digital audio watermarking method and apparatus using a color image as a watermark - Google Patents

Abstract

Color images of relatively large size are embedded in the audio signal data as watermarks. The audio signal data having the watermark already inserted therein can be reproduced with sound quality very close to that of the original audio signal. The original audio data is reconstructed into a matrix form corresponding to the matrix form of the color image file before wavelet transform, while the color image file is converted in format before discrete cosine transform. The transformed audio signal data and the coefficients of the reformatted color image are combined before inverse matrix reconstruction, and then inverse wavelet transformed, resulting in an audio signal embedded with the color image watermark.

Description

DIGITAL AUDIO WATERMARKING METHOD AND APPARATUS USING A COLOR IMAGE AS A WATERMARK}

The present invention relates to a technique for digitally watermarking data such as audio, video, and multimedia data. In particular, the present invention relates to a technique for inserting a watermark signal having a relatively large data size, such as a color image, into digital audio data.

As digitalized works such as images, videos and multimedia data proliferate, there is a need for a security system that can easily identify the source of such digital works. In particular, the use of the Internet to transmit recorded music in digital format is increasing. Thus, content providers, e.g., owners of the recorded music in the digital format described above, may be able to subsequently detect with software and / or hardware devices in order to authorize control, management, and copyright of multimedia data. Since it is necessary to insert a mark of M into multimedia data, digital watermarking has been developed as a technique of inserting identifiable data into multimedia data.

Conventionally, watermark signals used to watermark audio signals have been relatively simple signals such as a series of code codes. This is because, unlike inserting a watermark signal into an image or video, inserting a large watermark signal into an audio signal has an effect on intercepting the original audio. Moreover, since watermarks can be removed without authentication in conventional watermarking techniques, it becomes difficult to track the source of copyrighted works.

It is an object of the present invention to provide a digital audio watermarking method and apparatus which can insert relatively large color image data into audio signal data without compromising the quality of the audio signal when reproducing the audio signal. Thus, the watermarking embedding method according to the present invention provides a high correlation between the original audio signal and the watermarked version of the audio signal. This object is achieved by reconstructing the audio data structure to correspond to the image file to be inserted and performing a frequency domain transform.

1 is a block diagram illustrating a process of inserting a watermark signal into audio signal data according to the present invention;

FIG. 2 shows a process of the fast wavelet transform (FWT) group shown in FIG.

3 shows a process of an inverse fast wavelet transform (FWT) machine shown in FIG.

4 shows a color image watermark used in an experiment.

5 shows the spectrum of the original digital audio.

6 shows a spectrum of digital audio with a watermark embedded in accordance with the present invention;

Fig. 7 shows a color image watermark extracted from digital audio with a watermark embedded therein.

8 illustrates a distribution mechanism of watermarked audio data in accordance with the present invention.

<Explanation of symbols for the main parts of the drawings>

10: matrix structure reconstructor

20: Wavelet Converter

30: Color Image Format Converter

40: Discrete Cosine Transform (DCT)

50: Color Image Information Extractor

60: matrix coefficient adjuster

70: inverse matrix structure reconstructor

80: Inverse Wavelet Converter

1, the present invention will be described in detail.

A matrix structure reconstructor 10 is arranged to reconstruct the original audio digital data into the same matrix format as the color image to be inserted later as a watermark. The original audio data stream is divided into a number of sectors. At this time, the size of each sector is equal to the size of the image file. The image file is preferably inserted intermittently, but may be inserted in all sections of the audio data. The color image format converter 30 converts the original color watermark image file in a format such as BMP (bitmap) or JPEG (Joint Photographic Experts Group) to YIQ format. Where Y is luminescence, I is in-phase component, and Q is quadrature component. This conversion makes it possible to process large color images at high speed in a later step. In the embodiment of the present invention, only the Y component is used as an example. After this conversion, the scale, size, color information, bit structure of the converted image file is extracted by the color image information extractor 50 and, as described above, to reconstruct the original audio data using this information. The matrix structure reconstructor 10 is provided.

The reconstructed audio data and the YIQ converted image data are converted into the frequency domain by the wavelet converter 20 and the discrete cosine converter 40 (hereinafter referred to as DCT device), respectively. Fourier transform is also used to convert the watermark into the frequency domain to conceal the content of the watermark. In the Fourier transform, when an impulse-type watermark is converted, coefficients of the spectrum are spread throughout the spectrum, so that once the watermark is inserted into the target data, it is difficult to remove the inserted watermark. However, since the Fourier transform coefficients are complex, it is not easy to insert these coefficients into the target data. Therefore, it is preferable to perform DCT in the DCT group 40. This is because DCT is similar to the Fourier transform, especially the fast Fourier transform, but generates only coefficients that are real in the transformed region. Conventionally, DCT has been used to encode a signal. For example, the JPEG standard was used for data compression. Basically, DCT is defined as in Equation 1 below.

Where s is the original value of N and t is the converted value of N. In addition, the coefficient c is defined as follows.

Using the above two definitions, the two-dimensional DCT may be defined as in Equation 3 below.

Here, N, s and t represent the same parameters as those of the one-dimensional DCT, and c (i, j) is defined as in Equation 4 below.

Inverse DCTs defined in one and two dimensions are defined as follows, respectively.

To convert the reconstructed audio data, wavelet transform is used in wavelet converter 20. As the Fourier transform uses a sine and cosine function as the basis function, the wavelet transform uses a wavelet as the basis function. Wavelet transforms fall into two categories. One of them is a continuous wavelet transform, which is defined similarly to the Fourier transform and is defined as follows.

Extended wavelets exhibit low frequency components at high scale, while compressed wavelets exhibit high frequency components at low scale. Thus, scale and frequency are inversely related. Since the coefficients generated by the continuous wavelet transform are functions of scaling and position shift, the number of coefficients generated by the continuous wavelet transform is infinite. Thus, a discrete wavelet transform based on a subset of scaling and position movement is used. More effective algorithms can be provided by determining scaling and positional shifts according to dyadic. Nevertheless, a significant amount of computation is inevitable. To solve this drawback, fast wavelet transform is proposed. The fast wavelet transform technique employs a conventional two-channel subband coding and pyramid algorithm, and employs a so-called Perfectly Reconstruction Quadrature Mirror Filter (PR-QMF) based on the interconnection relationship between filter banks for inverse transform. It can be implemented relatively simply. 2 and 3 are diagrams illustrating processes of fast wavelet transform and inverse fast wavelet transform, respectively.

After the conversion as described above, in order to minimize the distortion of the original audio due to the watermark embedding, the respective coefficients of the thus-converted converted audio data and image have a scaling factor, for example, in the matrix coefficient adjuster 60. By multiplication. After this adjustment, the two sets of coefficients are summed together. Inverse matrix structure reconstructor 70 then reconstructs the summed coefficients such that the summed coefficients are the same as the structure of the original audio data. Finally, the reconstructed coefficients are inverse wavelet transformed by the inverse wavelet converter 80 to generate audio signal data with a color image embedded therein.

4 is a diagram showing a spectrum of an original audio signal, and FIG. 5 is a diagram showing a color image to be inserted into an audio signal in an experiment. 6 is a diagram illustrating a spectrum of an audio signal with a color image inserted in accordance with the present invention. As can be readily seen from these figures, the two spectra are almost identical, meaning that the original audio was not negatively affected. Correlation between two signals may be obtained using Equation 8 below widely used in the art.

Experiments show a 98.5% correlation, which is better than the 95% minimum correlation. This is because the present invention can reduce distortion of the original audio file while inserting a relatively large image file. Moreover, the extracted watermark has more than 75% correlation with the original watermark image. FIG. 7 is a diagram showing the watermark of FIG. 5 after being extracted from the signal with the watermark embedded therein. The extracted watermark can be easily recognized as the original watermark.

8 is a diagram illustrating one of the environments in which the present invention is used. As described above with reference to FIG. 1, watermarked audio generated from the watermarker 100 is distributed / downloaded on-line to remote computers 600 via an on-demand server 400. It may be stored in the database 200 to make. For this purpose, such an on-demand server is preferably connected to a communication network formed of a high-speed communication network such as B-ISDN or ATM-LAN, for example. Specifically, the communication network is an open network such as the Internet 500 or a network that uses a dedicated line such as personal computer communication. Thus, it is possible to use a remote computer or a terminal to connect to a server providing watermarked audio stored in a database. To verify that the extracted watermark is genuine, the watermark itself may be stored in another database 300 for download by an authorized individual.

In addition, the present invention is widely recognized and widely recognized in the field of digital image verification and protection, including the steps of verifying and protecting a digital image stored and stored in memory for the purpose of verifying and protecting the digital image. Can be used.

The present invention relates to determining whether digital audio / images have been modulated. In the technique of the present invention, information / watermarks are "stamping" into the original audio / multimedia data. This stamping technique ensures that the auditory / visual artifacts do not affect the original data.

Although a method and system for inserting color image watermark data by discrete cosine transforming a watermark signal and wavelet transforming the original audio data has been described and illustrated, those skilled in the art will only be aware of the scope of the appended claims. It will be apparent that modifications or variations of the present invention are possible without departing from the spirit and teachings of the invention.

According to the present invention in which a relatively large color image data is inserted into the audio signal data without damaging the quality of the audio signal when reproducing the audio signal, a general color image is used as the watermark signal, and the original data due to the watermark insertion is used. There is almost no distortion, freely inserting specific color images desired by individual users, not only numerical methods by correlation analysis, but also directly identifying images, and mixing and separating multimedia data of different formats. .

Claims (12)

In the method for embedding a color image watermark in audio data, Converting the color image watermark data from a first mode to a second mode, Reconstructing the audio data into a matrix corresponding to the matrix of the color image watermark data converted to the second mode; Generating a first spectral coefficient by wavelet transforming the reconstructed audio data into a frequency domain; Generating a second spectral coefficient by discrete cosine transforming the color image data of the second mode into a frequency domain; Combining the first and second spectral coefficients, and Converting the combined coefficients to generate audio data embedded with the color image watermark, The first mode is an RGB color mode representing pixels using red (R), green (G), and blue (B) components, and the second mode is a mode representing pixels using brightness components and color components. Characterized in that the method. The method of claim 1, Obtaining information regarding the color image watermark data; and Using the information in the reconstruction step. The method of claim 2, Wherein the information comprises one or more of the image size, the number of colors per pixel, and a bit array structure. The method of claim 1, The step of converting the combined coefficients to generate audio data in which the color image is inserted, Inversely reconstructing the combined coefficients into the original data structure of the audio data, and Inverse wavelet transforming the reconstructed combined coefficients to produce watermarked audio data. delete A digital watermarking apparatus for embedding a color image watermark in audio data, comprising: Means for converting the color image watermark data from a first mode to a second mode, Means for reconstructing the audio data into a matrix corresponding to the matrix of color image watermark data converted to the second mode; Means for wavelet transforming the reconstructed audio data into a frequency domain to generate a first coefficient; Means for discrete cosine transforming the color image data in the second mode to produce a second coefficient; Means for combining the first and second coefficients, and Means for transforming the combined coefficients to produce audio data embedded with the color image watermark, The first mode is an RGB color mode representing pixels using red (R), green (G), and blue (B) components, and the second mode is a mode representing pixels using brightness components and color components. Device characterized in that. The method of claim 6, Means for obtaining information regarding the color image watermark data, and Means for using the information in the means for reconstructing. The method of claim 7, wherein Wherein the information includes one or more of the image size, the number of colors per pixel, and a bit array structure. The method of claim 6, The means for transforming the combined coefficients to produce audio data with the color image embedded therein, Means for reconstructing the combined coefficients back into the original data structure of the audio data, and Means for inverse wavelet transforming the reconstructed combined coefficients to produce watermarked audio data. delete The method of claim 6, And storage means for storing the watermarked audio data and the color image watermark. The method of claim 11, And a server connected to said communication network for communicating said watermark embedded audio data stored in said storage means to a remote terminal connected to a communication network.