JP3812744B2 - Image alteration detection system - Google Patents

Description

  The present invention relates to, for example, an image alteration detection system that detects whether or not a produced image has been altered, and more particularly, to an image alteration detection system that effectively detects alteration of an image.

  In recent years, still images and moving images are often used as important evidence photographs. For example, identification of a criminal in an automatic payment machine such as a bank or a surveillance camera in a general store, and application to medical diagnosis using a medical image such as an X-ray photograph. On the other hand, with the digitization of image processing, it has become possible to intentionally modify and alter the original image without causing any deterioration in image quality from the original image. For this reason, being able to identify whether a given image is an original image or a modified image is extremely important in determining whether to adopt an image as evidence (for example, (See Patent Documents 1 and 2.)

There is an encryption technique as a conventional technique for preventing the original image from being modified in advance. In encryption technology, an original image is keyed so that only the person who owns the key can decrypt the original image. However, if the original image is decrypted by stealing the key by some illegal means, it is possible to tamper with the original image after that, and this is not an essential measure. In addition, it is inconvenient to use the key to see the original image.
In addition, as a conventional technique for detecting (detecting) whether or not the original image has been modified afterwards, the format of the image data is determined and some information describing the characteristics of the original image is added to the image data as a header, for example. There is technology to keep. However, if the image data and the header can be separated by some device before the alteration, it is possible to disable the detection of the alteration by altering the header in accordance with the alteration of the original image. It is not an appropriate measure.
In addition, as information describing the characteristics of the original image, for example, there is an example in which a Hash value obtained by a Hash function is used (see, for example, Patent Document 1). It may be difficult to obtain a Hash value for an image in real time.

JP 2003-289435 A JP 2002-271609 A Supervised by Eto and Kaneko, "Error correction codes and their applications", published by Ohm Koko Matsui, Journal of the Institute of Image Information and Television Engineers, April 2000, p. 565-566

As described above, conventionally, there has been a strong demand for the development of an effective system for detecting falsification of an image.
The present invention has been made in view of such a conventional situation, and an object thereof is to provide an image alteration detection system capable of effectively detecting whether or not an image has been altered.
More specifically, in the present invention, for example, whether or not information describing the characteristics of the original image is added somewhere in the image cannot be perceived only by viewing the image by a third party, and the image An image alteration detection system is realized in which the feature information added in advance remains without being lost even after the image is altered and it is difficult to intentionally rewrite the additional information. Specifically, in the present invention, for example, for a moving image, it is possible to detect alteration of an image using information describing characteristics of an original image that can be easily obtained in real time.

In order to achieve the above object, in the image alteration detection system according to the present invention, the post-processing device detects that the image data generated by the pre-processing device has been altered by the following configuration.
That is, in the preprocessing device, the preprocessing spatial frequency domain conversion means converts the spatial domain image data into the spatial frequency domain image data. The error-detectable code data generating means generates code data that can be detected using the remaining part of the image data in the spatial frequency domain transformed by the preprocessing spatial frequency domain transforming means except for the whole or a part thereof. The error-detectable code data adding means adds the error-detectable code data generated by the error-detectable code data generating means to the spatial frequency domain image data converted by the preprocessing spatial frequency domain converting means. The spatial domain converting means converts the image data to which the code data that can be detected by the error detectable code data adding means is added to the image data of the spatial domain.

  In the post-processing device, the post-processing spatial frequency domain conversion means converts the image data in the spatial region to which the error-detectable code data generated by the pre-processing device is added into image data in the spatial frequency region. The error detectable code data detection means detects code data capable of error detection added to the image data in the spatial frequency domain converted by the post-processing spatial frequency domain conversion means. There is an error in the image data in the spatial frequency domain transformed by the post-processing spatial frequency domain transforming means, using the error detectable code data detected by the error detectable code data detecting means by the image data alteration detection means. By detecting whether or not the image data has been tampered with.

Therefore, when image data of a spatial region in which code data capable of error detection is added in the spatial frequency region is generated by the preprocessing device and the image data of the spatial region is falsified, the image data of the spatial region is Since it is detected by the post-processing device that the image has been tampered with, it is possible to effectively detect whether or not the image has been tampered with by adding code data capable of error detection to the image data such as digital watermark information. can do.
Specifically, for example, whether or not information describing the characteristics of the original image is added somewhere in the image cannot be perceived only by viewing the image by a third party, and after the image has been tampered with Also, the feature information that has been added in advance remains without being lost, and image alteration detection that makes it difficult to intentionally rewrite the additional information can be realized. Specifically, for example, for a moving image, it is possible to detect alteration of an image using information describing characteristics of an original image that can be easily obtained in real time.

Here, various types of image data to be processed may be used. For example, still image data may be used, or moving image data may be used.
As the image data in the spatial domain, for example, image data on an XY orthogonal coordinate plane is used, and as the image data in the spatial frequency domain, for example, (spatial frequency in the X direction−in the Y direction). (Spatial frequency) image data on an orthogonal coordinate plane is used. Note that the characters X and Y are used for convenience of explanation, and are not particularly limited, and other characters may be used.

Also, code data that can detect errors using the entire image data in the spatial frequency domain, and code data that can detect errors using the remaining part of the image data in the spatial frequency domain are generated. For example, various modes may be used. For example, a mode in which code data that can detect an error is generated for each predetermined unit area constituting the image data in the spatial frequency domain may be used. Alternatively, a mode may be used in which code data capable of error detection is generated by collecting the entire image data in the spatial frequency domain.
Various parts may be used as the image data part used for generating each error-detectable code data.

In addition, various code data may be used as code data that can be detected by an error. For example, error detection code data can be used. In general, error detection code data also has error correction capability, and can be used as error correction code data.
Various modes may be used for adding error-detectable code data to image data in the spatial frequency domain, for example, adding error-detectable code data to a part of the image data. Thus, it is possible to use a mode in which the addition result is the partial data, or a mode in which a part of the image data is replaced with error-detectable code data.

Further, as a method of detecting code data capable of error detection added to the image data in the spatial frequency domain in the post-processing device, for example, various methods generally known for digital watermark technology are used. For example, a method of analyzing and detecting the image data or a method of receiving information on the addition state of code data that can be detected from a preprocessing device or the like and detecting based on the information is used. Can do.
Also, in the post-processing device, when it is detected that there is an error in the image data in the spatial frequency domain, it is usually detected that the image data has been tampered with, and there is no error in the image data in the spatial frequency domain. Is detected, it is detected that the image data has not been tampered with.

Note that the pre-processing device and the post-processing device may be configured separately, or may be configured integrally, for example.
The pre-processing device and the post-processing device may be configured as various devices or systems. For example, the pre-processing device and the post-processing device may be configured as an image processing device separately or integrally. Configuring the processing device as a wired or wireless transmission device and configuring the post-processing device as a wired or wireless receiving device, or wired or wireless communication having both the pre-processing device function and the post-processing device function A device can be configured.

The image alteration detection system according to the present invention has the following configuration as one configuration example.
That is, the error-detectable code data generating means of the pre-processing device uses the remaining part excluding a part of the spatial frequency domain image data converted by the pre-processing spatial frequency domain converting means to detect error data. Is generated. The error-detectable code data adding means of the preprocessing device is a spatial frequency domain image obtained by converting the error-detectable code data generated by the error-detectable code data generating means by the preprocessing spatial frequency domain converting means. The data is added to the partial data or replaced with the partial data.

Therefore, by adding error-detectable code data generated by using the remaining part excluding the part to a part of the image data in the spatial frequency domain, a good error as a whole can be obtained. Detection can be performed. For example, when adding code data capable of error detection generated from the remaining part (for example, most part) excluding the part to a part of the image data, the part of the image data and the part are By forming all the image data from the remaining portion, it is possible to use all of the image data for error detection without waste.
Here, various parts may be used as a part of the image data in the spatial frequency domain to which the error-detectable code data is added or replaced, and the remaining part excluding the part, for example, The high frequency component is used as the part, and all other components (low frequency components or components that are considered to be relatively low frequency) are used as the remaining part except the part. I can do it.

The image alteration detection system according to the present invention has the following configuration as one configuration example.
That is, the image data is moving image data composed of a plurality of frames of image data.
Further, the error detectable code data generating means and the error detectable code data adding means of the preprocessing device perform processing in a manner in which the image data of each frame is divided into a plurality of blocks.
The error-detectable code data adding means of the preprocessing device is a spatial frequency domain image obtained by converting the error-detectable code data generated by the error-detectable code data generating means by the preprocessing spatial frequency domain converting means. The data is added at a different position for each frame.

Therefore, error detection is performed in relatively small blocks by dividing the image data of each frame into a plurality of blocks and generating code data that can detect errors and adding code data that can detect errors. Can be realized.
Further, by adding error-detectable code data to different positions (components) in image data for each frame, for example, making error-detectable code data added to image data inconspicuous, It can be prevented from being decrypted by a third party.

Here, as an aspect of dividing the image data of each frame into a plurality of blocks, an aspect of generating code data that can detect errors with respect to a plurality of blocks, and an aspect of adding code data that can detect errors with respect to a plurality of blocks Various aspects may be used respectively.
In addition, various modes may be used for adding error-detectable code data to different positions for each frame. For example, the position where code data that can detect errors is added is a predetermined number. A pattern that repeats with the period of the frame can be used.

In the image alteration detection system according to the present invention, as one configuration example, in the post-processing device, the image data error correction means uses post-processing code data detected by the error-detectable code data detection means. An error included in the image data of the spatial frequency domain converted by the spatial frequency domain conversion means is corrected.
Therefore, it is possible to correct an error generated in image data due to falsification, transmission error, recording error, or the like, using code data that can detect an error and has error correction capability.

From here, instead of processing in the spatial frequency domain, processing in the spatial domain will be described.
In order to achieve the above object, in the image alteration detection system according to the present invention, the post-processing device detects that the image data generated by the pre-processing device has been altered by the following configuration.
That is, in the preprocessing device, the preprocessed image data partial identification means identifies part of data to be replaced with error-detectable code data and the remaining part of the image data in the spatial domain. Error-detectable code data generating means generates error-detectable code data using the remaining portion of data identified by the preprocessed image data part identifying means. The error detectable code data replacement means replaces the error detectable code data generated by the error detectable code data generation means with the partial data identified by the preprocessed image data partial identification means.

  In the post-processing device, the post-processing image data partial identification means replaces the error-detectable code data with respect to the image data in the spatial region to which the error-detectable code data generated by the pre-processing device is added. Identify some data and the rest of the data. The image data falsification presence / absence detection means is identified by the post-processing image data partial identification means using the error-detectable code data specified from the partial data identified by the post-processing image data partial identification means. By detecting whether or not there is an error in the remaining portion of the data, it is detected whether or not the image data in the space area has been tampered with.

  Therefore, part of the image data in the spatial domain is replaced with code data that can be detected by using the remaining part of the data other than the part of the data, and the code data that can be detected by the error is added. If the image data is generated by the pre-processing device and the image data is falsified, the falsification is detected by the post-processing device using the error-detectable code data included in the image data. By adding detectable code data to image data such as digital watermark information, it is possible to effectively detect whether or not the image has been tampered with.

Here, with respect to the image data in the spatial domain, various methods are used as a method of discriminating a part of the data to be replaced (or replaced) with code data capable of detecting an error and the remaining part of the data. For example, it is possible to use a method in which the part and the remaining part are set in advance and are identified based on the contents of the setting.
In addition, regarding the image data in the spatial domain, various data may be used as the partial data to be replaced with the error-detectable code data and the remaining data, for example, the partial data As the data, the least significant bit data of the predetermined pixel data constituting the image data can be used.
In addition, various modes are used as modes for generating error-detectable code data by using the remaining part of the data except for a part of the spatial domain image data that is replaced with error-detectable code data. Also good.

  Also, the configuration for adding code data capable of error detection to spatial domain image data is the same as that described for the configuration for adding code data capable of error detection to spatial frequency domain image data. It is possible to apply. For example, as a mode of adding error-detectable code data to image data, a mode of adding error-detectable code data and image data, using moving image data as image data, Processing image data by dividing it into multiple blocks, adding error-detectable code data to different positions in the image data for each frame, and correcting errors using error-detectable code data Etc. are possible.

  As described above, according to the image falsification detection system according to the present invention for processing in the spatial frequency domain, the preprocessing device converts the spatial domain image data into the spatial frequency domain image data, and Using the remaining part of the image data in the spatial frequency domain except the whole or a part thereof, code data capable of error detection is generated, and the generated code data capable of error detection is converted into the image data of the converted spatial frequency domain. The image data to which the error-detectable code data is added is converted into image data in the spatial domain, and in the post-processing device, the space to which the error-detectable code data generated by the pre-processing device is added The image data of the area is converted into the image data of the spatial frequency domain, and the code data that can be detected by the error is added to the converted image data of the spatial frequency domain. Detecting whether the image data has been tampered with by detecting whether there is an error in the converted spatial frequency domain image data using the detected error-detectable code data Thus, since the post-processing device detects that the image data generated by the pre-processing device has been tampered with, by adding error-detectable code data to the image data, such as digital watermark information, It is possible to effectively detect whether the image has been tampered with.

  Also, with regard to processing in the spatial domain, according to the image alteration detection system according to the present invention, the preprocessing device replaces part of the data in the spatial domain with code data that can be error-detected and the rest of the data. And generating code data capable of error detection using the identified remaining portion of data, and replacing the generated code data capable of error detection with the identified partial data. In the post-processing device, the partial data in which the error-detectable code data is replaced with the remaining portion of the data in the spatial domain image data to which the error-detectable code data generated by the pre-processing device is added. And the remaining portion of the identified data has an error using the code data that can be detected from the identified portion of the data. Since the post-processing device detects that the image data generated by the pre-processing device has been detected by detecting whether or not the image data of the space area has been tampered with By adding detectable code data to image data such as digital watermark information, it is possible to effectively detect whether or not the image has been tampered with.

Embodiments according to the present invention will be described with reference to the drawings.
In the image alteration detection system according to the present embodiment, an error detection code is used as information (feature information) describing characteristics of the original image (see, for example, Non-Patent Document 1), and the feature information is added to the original image. A digital watermark technique is used as the technique (see, for example, Non-Patent Document 2).

First, the digital watermark technique will be described in detail.
For example, Non-Patent Document 2 shows a typical technique for digital watermarking.
With reference to an example of a conceptual state of the digital watermark shown in FIGS. 6A to 6D, a representative method of the digital watermark will be described.
(1) As shown in FIG. 6A, an image at a spatial position (for example, on the XY orthogonal coordinate plane) is divided into blocks each consisting of 4 pixels in total of 2 pixels vertically and horizontally. Here, as shown, the amplitudes of these four pixels are I ₀₀ , I ₁₀ , I ₀₁ , and I ₁₁ . For I _mn, of the two subscripts m and n added to I, m represents the horizontal location in FIGS. 6A and 6D (the value increases from left to right). , N represents a place in the vertical direction (a numerical value increases from top to bottom).

(2) The amplitudes of these four pixels are expressed by components w ₀₀ , w ₁₀ , w ₀₁ , w of four spatial frequencies (for example, spatial frequency in the X direction−spatial frequency in the Y direction) according to Equations 1 to 4. ₁₁ is converted. An image of the spatial frequency obtained by the conversion is shown as shown in FIG. For _wmn, of the two subscripts m and n added to w, m represents the horizontal location in FIGS. 6B and 6C (the value increases from left to right). , N represents a place in the vertical direction (a numerical value increases from top to bottom).

Note that the inverse transformation from these four spatial frequency components w ₀₀ , w ₁₀ , w ₀₁ , and w ₁₁ to the four spatial position components I ₀₀ , I ₁₀ , I ₀₁ , and I ₁₁ is realized by Equations 5 to 8. The

Here, it is assumed that, for example, 1-bit watermark information is added to the block of 4 pixels shown in FIG. The amplitude of information is 0 for information of “0” value and is b for information of “1” value. Although the watermark information may be added to any component of the spatial frequency, here, the addition of the component w ₁₁ watermark effect is large.
In addition, as shown in Non-Patent Document 2, for example, an amplitude a is added to w ₀₁ and an amplitude c is added to w _10. Although it is possible to embed the watermark information bits, it is shown here when adding the amplitude b only w ₁₁ in order to simplify the description.
(3) That is, when the watermark information is “1”, the original spatial frequency w ₁₁ is transformed as shown in Equation 9 as shown in FIG.

(4) Using (w ₁₁ + b) shown in Equation 9 and w ₀₀ , w ₁₀ , and w ₀₁ , as shown in FIG. Converted as shown in

That is, when the watermark information is “0”, the original pixel value is completely restored, but when the watermark information is “1”, the component b is distributed to four pixels. In addition, since the polarity is opposite between adjacent pixels in both the vertical and horizontal directions, as long as the amplitude of b is small, it is difficult to detect it as an obstruction to the original image. That is, b is embedded in the image as watermark information so as not to be noticed by a third party.
Meanwhile, for the parties know that the w ₁₁ components might watermark information is embedded, from the values of the four pixels of the image obtained as shown in FIG. 6 (D), calculation of formula 14 (W ₁₁ + b) is obtained. The presence of b can be identified by setting the value of w ₁₁ that the image originally has as shown in Expression 15.

Actually, in any case, if the value of b is set so as to satisfy w ₁₁ <b, b may become large, and the presence of watermark information may be known only by looking at the image. is there. For this reason, as a method for setting b as small as possible, there is a method for determining the existence of b by embedding the same information b in several blocks and taking the majority of identification results for each block.
In addition, there is a technique in which watermark information is not embedded uniformly in an original image but is embedded with emphasis in an outline portion of an image that is difficult to see as watermark information. In the above description, an example of a still image is shown. However, since a moving image can be interpreted as a group of still images that change continuously, the same technique as that for a still image can be applied to the moving image. It is known that in a moving image, it is effective to embed watermark information in a portion having a motion in the image with emphasis.

In addition, a method of converting a pixel value at a spatial position into a spatial frequency value is generally called orthogonal transformation, and for example, discrete Fourier transform (DFT), discrete cosine transform (DCT), Wavelet transform, or the like is used. Used.
In the above example, four pixels each having two vertical and horizontal pixels are used as one block unit. However, in DCT frequently used for image data compression such as JPEG and MPEG, vertical and horizontal eight pixels are used. There are many examples in which 64 pixels each are made into one block. In JPEG and MPEG, since DCT is used as one of image data compression methods, watermark information can be embedded using existing DCT without performing DCT for embedding watermark information.

Next, the error detection code will be described in detail.
An error detection code is generated in a total (A + B) word (or bit) by generating an error detection code of B word (or bit) for the original data of A word (or bit). The presence or absence of an error is detected (see, for example, Non-Patent Document 1).
As an example, let us consider a case in which a 2-bit error detection code is added to 8-bit data to obtain 10-bit data as a whole.
In this case, since only 256 types of values that can be expressed by 8 bits are used from among 1024 types of values that can be expressed by 10 bits, the remaining 768 types of values cannot originally exist as data. By utilizing this property, even if a 2-bit error occurs in 10 bits, it can be reliably detected.

Although there are various types of specific error detection codes, Reed-Solomon codes suitable for word unit processing are frequently used for systems such as images in which the concept of pixels (words) is clear. Yes. For example, VTR (Video Tape Recorder) and DVD (Digital Versatile Disk) which are digital recording of images are good examples.
Here, in the application example of the conventional error detection code, the cause of the error is random noise or the like, the probability that a small amount of error occurs is high, and the probability that a large amount of error occurs simultaneously is low. On the other hand, when the image alteration is interpreted as a code error as in this embodiment, it may be assumed that all pixel data is erroneous in an extreme case. The effect can be exerted even in such cases.

In addition, as described above, a 2-bit error detection code added to 8 bits can detect 2-bit errors in 10 bits, but a 1-bit error generated in 10 bits by the same code. It is known that (that is, half of detectable errors) can be corrected. In this way, error detection and error correction are essentially based on the same principle, and whether the function of the code is used for error detection or error correction is the example of image recording as described above. Then, it depends on the processing method of the reproduced image.
For this reason, in this embodiment, description will be made by referring to an error detection code (a code having an error detection function), but even if referred to as an error correction code (a code having an error correction function), the same or similar function is included. It is a waste.
In addition, operations such as error detection code generation and error detection and correction are realized in real time even for high-speed moving images such as high-definition television (HDTV). The technology to do this is established in the above-mentioned VTR and DVD.

An image alteration detection system according to a first embodiment of the present invention will be described.
The image alteration detection system of the present example includes a pre-processing device that adds digital watermark information including an error detection code to an image, and a post-processing device that reproduces the original image from the image added with the digital watermark information including an error detection code. It is configured. The pre-processing device and the post-processing device may be integrated or may be separate. For example, the image processed by the pre-processing device may be recorded on a recording medium, and the post-processing device may read the image from the recording medium, or the image processed by the pre-processing device may be transmitted to the post-processing device. it can.
FIG. 1 shows a configuration example of the pre-processing apparatus of this example. In this example, in order to simplify the description, a case where a black and white still image is first taken by the camera 1 will be described as an example.
The pre-processing device of this example includes a camera 1 that captures an image a1, a spatial frequency converter 2 that converts a component (I component) of an amplitude of a pixel in an image at a spatial position into a component (w component) of a spatial frequency, An error detection encoder 3 that generates an error detection code, an adder 4 that adds codes, and a spatial frequency that converts a spatial frequency component (w component) into a pixel amplitude component (I component) in an image at a spatial position An inverse converter 5 is provided.

An example of the operation performed by the pre-processing apparatus of this example will be shown.
An image a1 is taken by the camera 1, and an A / D (Analog to Digital) converter built in the camera 1 is used to output the image from the camera 1 as a digital value. For example, digital original image data a2 (for example, FIG. 6 (A) shown in FIG. 6A is sampled into M pixels in the horizontal direction and N pixels in the vertical direction, and the pixels at each sampling point are binary-coded with amplitude of K bits. )). These {(M × N) pixels × K} -bit image data are converted from the I component to the w component by the spatial frequency converter 2 (for example, an image as shown in FIG. 6B).
The w component includes, for example, a low frequency component a3 composed of w ₀₀ , w ₁₀ , and w ₀₁ and a high frequency component a4 composed of w ₁₁ . The low frequency component a3 is input to the error detection encoder 3, and the error detection encoder 3 generates an error detection code a5 based on the low frequency component a3. The error detection code a5 is added to the high frequency component a4 by the adder 4, and the addition result and the low frequency component a3 are input to the spatial frequency inverse transformer 5 (for example, an image as shown in FIG. 6C). ). In the spatial frequency inverse transformer 5, a digital watermarked image a6 in which the error detection code a5 is inserted into the original image data a2 as a digital watermark is generated from the addition result and the low frequency component a3 (for example, FIG. 6). Image as shown in (D)).

Note that even if the image a6 with digital watermark is viewed, the third party cannot recognize whether or not an error detection code is inserted as a digital watermark due to the effect of the digital watermark technique described above. It only looks as an original image.
Thereafter, the digital watermarked image a6 is stored in, for example, a storage device or transmitted by a transmission device.
Further, there is a case where the digital watermarked image a6 is intentionally altered during accumulation or transmission. Here, the alteration may be a spatial structural alteration such as deleting a part of the original image, replacing it with another new image, scaling, rotating, or blurring the image. Such as tampering, amplitude tampering such as making the image white or black. Any of these tampering results in amplitude changes in the frequency components of all or any of the pixels in the original image consisting of (M × N) pixels. That is, the same phenomenon as when a code error occurs in the frequency component constituting the original image data a2 is brought about.

FIG. 2 shows a configuration example of the post-processing apparatus of this example. The post-processing apparatus of this example uses the property of the digital watermarked image a6 of this example as described above.
The post-processing apparatus of this example includes a spatial frequency converter 11 that converts a pixel amplitude component (I component) in a spatial position image into a spatial frequency component (w component), and an identifier that identifies an error detection code b4. 12 and an error detection decoder 13 for performing error detection decoding.
An example of the operation performed by the post-processing apparatus of this example is shown.
The received image data b1 that may have been altered from the original image data a2 is converted again from the I component to the w component by the spatial frequency converter 11 and separated into the high frequency component b2 and the low frequency component b3. The high frequency component b2 includes an error detection code in addition to the high frequency component w of the original image, that is, the sum of the high frequency component w of the original image and the error detection code.
The discriminator 12 discriminates the error detection code b4 from the high frequency component b2 of the image data. Here, if there is no falsification, the low frequency component b3 and the error detection code b4 match the low frequency component a3 and the error detection code a5 of the original image, but if there is falsification, some change equivalent to a code error will occur. Has occurred.
The low frequency component b3 and the error detection code b4 are input to the error detection decoder 13, and the error detection decoder 13 outputs a detection result (falsification detection result) b5 indicating whether there is an error in the received image data b1, that is, whether the falsification has occurred. To do.

FIGS. 3A to 3C show more realistic numerical values of blocks as a mode different from the numerical values shown in FIGS. 6A to 6D.
In this example, as shown in FIG. 3A, an example is an image in which one frame is composed of horizontal 640 pixels and vertical 480 pixels. This is divided into 10 horizontal divisions and 15 vertical divisions to obtain an error detection block consisting of 64 horizontal pixels and 32 vertical pixels. An 8-word error detection code is generated for 2048 pixels (words) included in the error detection block.
Next, as shown in FIG. 3B, consider an intermediate block of 16 pixels by dividing the error detection block into four horizontal parts and two vertical parts. A one-word error detection code is inserted per intermediate block.
Further, as shown in FIG. 3C, when the intermediate block is divided into two horizontally and vertically, a basic block of 8 horizontal pixels and 8 vertical pixels is obtained.
Here, if the pixel is 8 bits, that is, if one word of the error detection code is also 8 bits, a 2-bit error detection code may be inserted per basic block.

In principle example of the electronic watermark as shown in FIG. 6 (A) ~ (D) , has been inserted one bit spatial frequency component w ₁₁ for two pixels square base block, in this example, 8 pixels square basic Insert 2 bits into the block. For this purpose, for example, w ₇₇ and w ₆₆ are selected as spatial frequency components into which 2 bits are inserted, and these are handled as the high frequency component a4. At the same time, for example, all components other than w ₇₇ and w ₆₆ are handled as the low frequency component a3.
In order to reliably identify the error detection code, the w ₇₇ and w ₆₆ components of the image are removed in advance, and only the error detection code a5 is regarded as the w ₇₇ and w ₆₆ components. It is also possible. However, in this configuration, the original image is deteriorated to some extent with the removal of the components w ₇₇ and w ₆₆ .

As described above, in this example, 64 bits, that is, an 8-word error detection code a5 is inserted using 32 sets of w ₇₇ and w ₆₆ in 32 basic blocks included in one error detection block. Can do.
In the post-processing, contrary to the above-described pre-processing, an 8-word error detection code b4 is extracted from one error detection block, and a code error is generated in 2048 pixels of the one error detection block using the error detection code b4. Is determined in the frequency domain. In this example, one frame of the image consists of 150 error detection blocks. If one of these blocks contains an error, it is determined that the image has been tampered with. For this determination, for example, the sum (OR) of these 150 types of error detection results may be taken.
In this example, since an error detection code of 8 words is used, tampering up to any 32 basic blocks in 2048 pixels of one error detection block can be reliably detected. For tampering of 32 basic blocks or more, there is a possibility that the tampering result matches an error-free code and the tampering may be overlooked. The probability of this oversight is 2 to the 8th power, that is, 1/256 (= Furthermore, there is a case where the alteration in the screen exceeds the range of one error detection block, and the probability of this oversight is further reduced by taking the sum (OR) of the alteration results.

The difficulty in perceiving digital watermarks in images, the certainty of extracting error detection codes embedded as digital watermarks, and the low detection oversight in the event of tampering are in a trade-off relationship. Depending on which one is emphasized, the above parameters can have various combinations.
In this example, a monochrome still image is taken as an example, but the same configuration and operation can be applied to a moving image. In the case of a color image, it is effective to embed digital watermark information in the luminance signal, for example, among the luminance signal and the two color difference signals constituting the color signal.
In this example, for example, the digital watermark information can be made invisible by shifting the spatial frequency component in which the digital watermark information is embedded between adjacent frames. In addition, this makes it possible to obtain an effect that the digital watermark information embedding algorithm is difficult to be seen by a third party. For example, without fixing the spatial frequency component for embedding digital watermark information to the components w ₇₇ and w ₆₆ , the next frame is the components w ₇₇ and w ₇₆ , the next frame is the components w ₇₇ and w ₆₇ , etc. As described above, a configuration may be adopted in which arbitrary two components are periodically selected from the high-frequency component a4 to be the target of embedding digital watermark information.

In the preprocessing apparatus of the image alteration detection system of this example shown in FIG. 1, the preprocessing spatial frequency domain transforming means is configured by the function of the spatial frequency converter 2, and an error is detected by the function of the error detection encoder 3. Detectable code data generating means is configured, error detecting code data adding means is configured by the function of the adder 4, and space domain converting means is configured by the function of the spatial frequency inverse transformer 5.
Further, in the post-processing apparatus of the image alteration detection system of this example shown in FIG. 2, post-processing spatial frequency domain conversion means is configured by the function of the spatial frequency converter 11, and an error can be detected by the function of the classifier 12. The code data detection means is configured, and the function of the error detection decoder 13 constitutes image data alteration presence / absence detection means.

An image alteration detection system according to a second embodiment of the present invention will be described.
In the first embodiment described above, the digital watermark information is embedded in the frequency domain of the image. In this example, as another configuration example, a configuration example in which the least significant bit representing the amplitude of the pixel is replaced with the digital watermark information is shown.
For example, in the case shown in FIGS. 3A to 3C, error detection information of 1 word (8 bits in this example) is embedded in an intermediate block of 16 pixels square shown in FIG. 3B. Among the 64 pixels of the 8 pixel square basic block shown in FIG. 3C, the least significant bit of the specific 2 pixels is replaced with 2 bits of the error detection code.

As an example, the least significant bit in I ₀₀ and I ₀₄ which is the amplitude of the pixel is to be replaced. Even with such replacement, the digital watermark information in the image cannot be detected. That is, for 62 pixels out of 64 pixels, the image is expressed using all 8 bits in one word, and only the remaining 2 pixels are expressed in the upper 7 bits. The deterioration of the image due to the disappearance of the low-order bits is slight, and even if the least significant bit in the two pixels becomes digital watermark information irrelevant to the image, the generation of noise is also small.
The positions of the two pixels used for embedding the digital watermark information in 64 pixels are I ₀₀ and I ₀₄ in the first frame, I ₁₀ and I ₁₄ in the next frame, and I ₂₀ and I ₁₄ in the next frame. and _24, so that the _{I 73, I 77} with 32 th frame, are shifted every frame, 32 frame periods in situ _{I 00,} when to return to _{I 04,} further difficult to detect the digital watermark information be able to. Specifically, in order from the first frame to 32 frames, (I ₀₀ , I ₀₄ ), (I ₁₀ , I ₁₄ ), (I ₂₀ , I ₂₄ ),..., (I ₇₀ , I ₇₄ ), (I ₀₁ , I ₀₅ ), (I ₁₁ , I ₁₅ ), ..., (I ₀₂ , I ₀₆ ), (I ₁₂ , I ₁₆ ), ..., (I ₀₃ , I ₀₇ ), ... The order of (I ₇₃ , I ₇₇ ) and then back to (I ₀₀ , I ₀₄ ) in the new first frame can be used.

With reference to FIG.4 and FIG.5, the pre-processing apparatus and post-processing apparatus which comprise the image alteration detection system of this example are demonstrated.
In this example, differences from the case of the first embodiment will be described in detail, and descriptions of similar points will be omitted or simplified.
Further, in this example, the same components as those shown in FIGS. 1 and 2 of the first embodiment are denoted by the same reference numerals.
FIG. 4 shows a configuration example of the preprocessing apparatus of this example.
The preprocessing apparatus of this example includes a camera 1, a watermark insertion position detector 21 that detects (separates) a position where digital watermark information is inserted in an image, an error detection encoder 3, and a switch 22.

An example of the operation performed by the pre-processing apparatus of this example will be shown.
The digital original image data a2 obtained by the camera 1 that captures the image a1 is converted by the watermark insertion position detector 21 into the bit of the position at which the digital watermark information is inserted (in this example, the maximum of predetermined two pixels in the basic block). It is separated into lower order bits c2 and other bits (bits that are not the insertion position of digital watermark information) c1. In the error detection encoder 3, an error detection code c3 is generated using the other bit c1. In this example, an error detection code c3 for 8 pixels is obtained for 2048 pixels of the error detection block. The error detection code c3 is replaced with the image bit at the insertion position of the digital watermark information by turning on / off the switch 22. All of the bits of the replaced error detection code c3 (a part of the lower bits) and the image bits that are not the insertion position of the digital watermark information (all the upper bits and the remaining lower bits) are It becomes an image (image with digital watermark) c4 in which digital watermark information is embedded.

FIG. 5 shows a configuration example of the post-processing apparatus of this example.
The post-processing apparatus of this example includes a watermark insertion position detector 31 that detects (separates) a position where digital watermark information is inserted in an image, and an error detection decoder 13.
An example of the operation performed by the post-processing apparatus of this example is shown.
An image (received image data) d1 that may have been tampered with is detected by the watermark insertion position detector 31 that detects the position where the digital watermark information is embedded, and the bit d2 that is not the insertion position of the digital watermark information and the digital watermark information. It is separated into bits (in this example, bits of error detection code) d3 which is the insertion position. The error detection decoder 13 regards the bit d3 that is the insertion position of the digital watermark information as an error detection code, detects a code error in the bit d2 that is not the insertion position of the digital watermark information, and falsifies the detection result. Output as detection result d4.

In this example, for example, by shifting the pixel position where the digital watermark information is embedded between adjacent frames so as to make a round in 32 frames, the digital watermark information can be made invisible. In addition, this makes it possible to obtain an effect that the digital watermark information embedding algorithm is difficult to be seen by a third party.
In the configuration of this example, compared to the configuration shown in FIG. 1 and FIG. 2 of the first embodiment described above, there is no conversion to the spatial frequency of the image and no inverse conversion from the spatial frequency. For example, as long as the insertion position of the digital watermark information is known, there is a possibility that the process of making it impossible to detect even if falsified is relatively easy. Further, regarding the details, in the configuration shown in FIGS. 1 and 2 of the first embodiment described above, when the bit of the error detection code a5 is 0, an image with digital watermark information is completely an original image. However, in the configuration of this example, the least significant bit at the position where the error detection code c3 is inserted is lost regardless of the value of the error detection code c3. Compared to the above, there is slight image quality degradation.

  In the case shown in FIG. 3, one frame of an image is divided into a plurality of intermediate blocks and then divided into a plurality of basic blocks, and an error detection code is generated for each basic block. Various image regions may be used to generate the image, and as an example, it may be configured to generate one error detection code for all pixels of one frame of the image. It is. In this configuration, instead of determining the presence / absence of the entire alteration by the sum (OR) of the alteration results for each block as in this example, the error detection result by one error detection code is the alteration detection result of the entire image. Corresponding to

In the preprocessing apparatus of the image alteration detection system of this example shown in FIG. 4, the preprocessed image data partial identification means is configured by the function of the watermark insertion position detector 21, and the function of the error detection encoder 3 An error-detectable code data generating unit is configured, and an error-detectable code data replacing unit is configured by the function of the switch 22.
Further, in the post-processing apparatus of the image alteration detection system of this example shown in FIG. 5, post-processing image data partial identification means is configured by the function of the watermark insertion position detector 31, and by the function of the error detection decoder 13. Image data falsification presence / absence detection means is configured.

An image alteration detection system according to a third embodiment of the present invention will be described.
As described above, the error detection code and the error correction code are codes based on essentially the same principle. In the image alteration detection system as shown in the first embodiment and the second embodiment described above, it is most important to detect “falsification” of an image, that is, “error” of the image. However, for example, in the case shown in FIG. 3, if up to 16 basic blocks in 2048 pixels of one error detection block are tampered with, this can be corrected.
Thus, in the image alteration detection system, not only detecting “falsification” of an image but also automatically correcting the altered image to the original image can be an advantage depending on the application. For example, it is possible to correct minor tampering other than intentional tampering, or a small amount of code error generated during image transmission or storage, by such correction processing. .

Therefore, in this example, in the configuration shown in FIGS. 1 and 2 of the first embodiment described above and the configuration shown in FIGS. 4 and 5 of the second embodiment described above, the post-processing device further receives signals. If the image data is data of an altered image (including an image in which an error has occurred), the image is altered (error) within the range of error correction capability of the error detection code inserted into the image. Equipped with an error correction function to correct Specifically, as an example, in the post-processing device shown in FIG. 2 or FIG. 5, a configuration in which the error detection decoder 13 is changed to an error correction decoder can be used. In this case, from the error correction decoder As the output, one or both of two systems such as “falsification detection result” and “falsification corrected image” can be outputted.
In this example, the image data error correction means is configured by the function of the error correction decoder provided in the post-processing apparatus shown in FIGS.

  As shown in the above-described embodiments (first embodiment to third embodiment), in the image alteration detection system according to an embodiment of the present invention, for example, a spatial region of a digitized image (for example, FIG. 6). In a region as shown in (A)) or a spatial frequency region (for example, a region as shown in FIG. 6B), a code having an error detection function for most information included in the region ( Error detection code) and replacing or superimposing the remaining part of the information included in the region with the remaining part of the information, the code having the error detection function is added to the original image as digital watermark information. When a code error is detected using the embedded code having the error detection function, it is detected that the image has been tampered with.

  Further, in the image alteration detection system according to an embodiment of the present invention, for example, a digitized image is divided into a plurality of spatial blocks, and the spatial region or spatial frequency region for each spatial block is divided into the region. The error detection function is formed by forming a code having an error detection function for most of the included information and replacing or superimposing the remaining partial information included in the region with the remaining partial information. When a code error is detected in at least one of the plurality of spatial blocks using the embedded code having an error detection function. Detects that the image has been tampered with.

  In addition, the image alteration detection system according to an embodiment of the present invention has an error detection function for most information included in a spatial region or a spatial frequency region of a digitized moving image, for example. A code having the error detection function is embedded in the original image as digital watermark information by forming a code and replacing or superimposing the remaining partial information included in the region. Thus, when a code error is detected using the embedded code having the error detection function, it is detected that the image has been tampered with. As a preferred configuration example, the position (component) of the spatial region or spatial frequency region in which the code having the error detection function is embedded is made different for each adjacent frame.

  In addition, in the image alteration detection system according to an embodiment of the present invention, for example, in a spatial region or a spatial frequency region of a digitized image, a code having an error detection function for most information included in the region. By embedding a code having the error detection function as digital watermark information in the original image by replacing or superimposing the remaining partial information included in the region, When a code error is detected using the embedded code having the error detection function, it is detected that the image has been tampered with, and at the same time, the tampering in the image is corrected.

Therefore, in the image alteration detection system according to the embodiment of the present invention, an error detection code for detecting a change in the state of the image is embedded in the image as digital watermark information, so that the existence of the error detection code is known to a third party. Therefore, it is possible to detect image alteration.
Specifically, for example, whether or not information describing the characteristics of the original image (in this example, an error detection code) is added somewhere in the image cannot be perceived only by viewing the image by a third party. In addition, even after the image has been tampered with, the feature information added in advance (in this example, the entire information including the error detection code for performing error detection) remains without being lost, and the intent In particular, it is possible to realize an image alteration detection system in which it is difficult to rewrite additional information (in this example, the entire information including an error detection code for error detection). More specifically, for example, also for moving images, detection of image alteration using information (in this example, an error detection code) describing characteristics of an original image that can be easily obtained in real time. Can be realized.

  As described above, in the image alteration detection system according to the embodiment of the present invention, an error detection code is used as a criterion for image alteration, and the error detection code is added to the image as digital watermark information. The error detection code may be generated from various components in the image. For example, in the case of the first embodiment described above, the error detection code is generated from a component having a low spatial frequency, and the second described above. In the case of the embodiment, it is generated from the upper bits of the luminance value (amplitude) of the pixel at the spatial position. Various methods may be used for inserting digital watermark information. For example, in the case of the first embodiment described above, digital watermark information is added to a component having a high spatial frequency. In the case of the second embodiment, it is added as a lower bit of the luminance value (amplitude) of the pixel at the spatial position. Further, for example, it is possible to divide an image into a plurality of blocks, and to generate or add an error detection code for each block, and to add an error detection code for each continuous frame. It is possible to change places.

In the image alteration detection system according to the embodiment of the present invention, for example, an error in the image can be detected using an error detection code inserted into the image, and the error in the image can be corrected. Specifically, it is possible to correct errors that occur when an image is transmitted through a communication channel, errors that occur when an image is recorded, and the like.
In addition, for example, compared with an error detection code, a complicated calculation is required when the Hash function is used, and the error detection code has a higher error detection performance than the Hash function for some errors (tampering). There is also an advantage such as high.

Here, the configurations of the image alteration detection system, the pre-processing device, the post-processing device, and the like according to the present invention are not necessarily limited to those described above, and various configurations may be used. The present invention can also be provided as, for example, a method or method for executing the processing according to the present invention, a program for realizing such a method or method, or a recording medium for recording the program. It is also possible to provide various devices and systems.
The application field of the present invention is not necessarily limited to the above-described fields, and the present invention can be applied to various fields.
In this embodiment, the apparatus and system for detecting or correcting image alteration are mainly described. However, an apparatus or system for detecting or correcting an error generated in an image may be implemented by a similar configuration and operation. Is possible.

In addition, as various processes performed in the image alteration detection system, the pretreatment device, the post-processing device, and the like according to the present invention, the processor stores, for example, a ROM (Read Only Memory) in a hardware resource including a processor and a memory. A configuration controlled by executing the control program may be used, and for example, each functional unit for executing the processing may be configured as an independent hardware circuit.
Further, the present invention can be grasped as a computer-readable recording medium such as a floppy (registered trademark) disk or a CD (Compact Disc) -ROM storing the control program, or the program (itself). The processing according to the present invention can be performed by inputting the program from the recording medium to the computer and causing the processor to execute the program.

It is a figure which shows the structural example of the pre-processing apparatus of the image alteration detection system which concerns on 1st Example of this invention. It is a figure which shows the structural example of the post-processing apparatus of the image alteration detection system which concerns on 1st Example of this invention. It is a figure for demonstrating an example of the embedding process to the image of an error detection code, (A) is a figure which shows an example of the image of 1 frame, (B) is a figure which shows an example of the image of an intermediate | middle block. Yes, (C) is a diagram showing an example of an image of a basic block. It is a figure which shows the structural example of the pre-processing apparatus of the image alteration detection system which concerns on 2nd Example of this invention. It is a figure which shows the structural example of the post-processing apparatus of the image alteration detection system which concerns on 2nd Example of this invention. It is a figure for demonstrating an example of the embedding process to the image of a digital watermark, (A) is a figure which shows an example of the image of a spatial position, (B) is a figure which shows an example of the image of a spatial frequency. (C) is a figure which shows an example of the image of the spatial frequency with which the watermark was embedded, (D) is a figure which shows an example of the image of the spatial position where the watermark was embedded.

Explanation of symbols

  1 ·· Camera, 2 · 11 ·· Spatial frequency converter 3 ·· Error detection encoder 4 ·· Adder 5 ·· Inverse spatial frequency converter 12 ·· Identifier 13 ·· Error detection decoding 21, 31... Watermark insertion position detector 22, switch a 1, image, a 2, original image data, a 3, b 3, low frequency component, a 4, b 2, high frequency component, a 5, b 4 C3, error detection code, a6, c4, digital watermarked image, b1, d1, received image data, b5, d4, falsification detection result, c1, d2, bits not at insertion position, c2, d3 .. bit at the insertion position,

Claims (5)

An image falsification detection system that detects by a post-processing device that image data generated by a pre-processing device has been falsified,
The preprocessing device includes preprocessing spatial frequency domain conversion means for converting spatial domain image data into spatial frequency domain image data;
An error that generates code data that can be detected by using all of the remaining high frequency components of the image data in the spatial frequency domain converted by the preprocessing spatial frequency domain conversion means. Detectable code data generating means;
Error-detectable code data adding means for adding the error-detectable code data generated by the error-detectable code data generating means to the part of the spatial frequency domain image data converted by the preprocessing spatial frequency domain converting means; ,
A spatial domain conversion unit that converts the image data to which the code data that can be detected by the error detection code data addition unit is added to the spatial domain image data,
The post-processing device includes post-processing spatial frequency domain conversion means for converting the spatial domain image data to which the error-detectable code data generated by the pre-processing device is added into spatial frequency domain image data;
Error-detectable code data detecting means for detecting error-detectable code data added to the part of the image data in the spatial frequency domain transformed by the post-processing spatial frequency domain transforming means;
By detecting whether there is an error in the spatial frequency domain image data transformed by the post-processing spatial frequency domain transforming means using the error detectable code data detected by the error detectable code data detecting means Image data falsification presence / absence detection means for detecting whether the image data has been falsified,
An image alteration detection system characterized by that. The image alteration detection system according to claim 1,
Before error detectable code data adding means of the processing device, the image data of the transformed spatial frequency domain by pre-treatment the spatial frequency domain transforming means an error detection code capable of data generated by the error detection can code data generating means replaced with adding a portion of the data or of the partial data,
An image alteration detection system characterized by that. In the image alteration detection system according to claim 1 or 2,
The image data is moving image data composed of a plurality of frames of image data,
The error-detectable code data generating means and the error-detectable code data adding means of the preprocessing device perform processing in a manner in which the image data of each frame is divided into a plurality of blocks,
The error-detectable code data adding means of the preprocessing device is a spatial frequency domain image obtained by converting the error-detectable code data generated by the error-detectable code data generating means by the preprocessing spatial frequency domain converting means. Add to the data at a different position for each frame,
An image alteration detection system characterized by that. The image alteration detection system according to any one of claims 1 to 3,
The post-processing device corrects an error included in the spatial frequency domain image data converted by the post-processing spatial frequency domain transforming means using the error detectable code data detected by the error detectable code data detecting means. Data error correction means ;
Means for outputting both the falsification detection result and the image data after falsification correction in which the error is corrected by the image data error correction means ,
An image alteration detection system characterized by that. An image falsification detection system that detects by a post-processing device that image data generated by a pre-processing device has been falsified,
The preprocessing device performs preprocessing for identifying the least significant bit of the data representing the amplitude of a part of pixels to be replaced with error-detectable code data for the image data in the spatial domain and all the other parts of the remaining bits. Image data partial identification means;
Error-detectable code data generating means for generating code data capable of error detection using the bits of all the remaining parts identified by the preprocessed image data part identifying means;
Error-detectable code for replacing error-detectable code data generated by the error-detectable code data generating means with the least significant bit of the data representing the amplitude of the partial pixel identified by the preprocessed image data partial identifying means Data replacement means,
The post-processing device includes a part of the bit data in which the error-detectable code data is replaced with respect to the image data in the spatial region to which the error-detectable code data generated by the pre-processing device is added, and all the other remaining bits . Post-processing image data part identifying means for identifying the bits of the part;
Using an error detection code capable of data specified from the identified said some bits workup image data portion identification means, the bits in all parts of the remaining identified by the post-processing image data portion identification means Image data falsification presence / absence detecting means for detecting whether or not the image data of the space area has been falsified by detecting whether or not there is an error ,
The image data is moving image data composed of a plurality of frames of image data,
The pre-processing device and the post-processing device perform processing in a manner in which the image data of each frame is divided into a plurality of blocks, and the position of the pixel that replaces the error-detectable code data with the least significant bit is between adjacent frames. Shift,
An image alteration detection system characterized by that.

Images