Dct and wiener filter based light response non-uniformity noise anonymization method and system

Abstract

The present disclosure provides a DCT and Wiener filter-based photoresponse non-uniformity noise anonymization method and system, relating to the technical field of information security, which comprises obtaining an image to be anonymized and dividing the image into blocks; performing discrete cosine transform on each block of the divided image, separating the high-frequency and low-frequency regions of each block of the divided image, extracting the high-frequency components of each block of the divided image and retaining the low-frequency components, performing multiple Wiener filter processing on the high-frequency components, merging the high-frequency components after the multiple Wiener filter processing and the retained low-frequency components, and restoring the merged whole into a preliminary anonymized divided image by using inverse discrete cosine transform; determining whether each block of the divided image has been restored into a preliminary anonymized divided image, and if so, merging all the preliminary anonymized divided images and obtaining a final anonymized image by using multiple Wiener filter processing on the merged whole. The present disclosure ensures high-quality images under the premise of effectively removing PRNU.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of information security, in particular to a photo-response non-uniformity noise anonymization method and system based on DCT and Wiener filtering. BACKGROUND

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute the prior art.

[0003] It is a widely used multimedia forensics method to trace the source camera by using the photo-response non-uniformity (PRNU) of a photosensitive chip, which can help forensic scientists to make forensic tests. At present, the rapid development of image technology has brought a series of privacy disclosure problems. From an image, not only can the shooting content, shooting time and shooting location be extracted, but also the fingerprint in the image can be extracted using the PRNU technology, so as to trace the source camera of the image and even find the photographer. Therefore, when the public uploads the photographed image to the network, it is very likely to attract the attention of unscrupulous people, resulting in the disclosure of the photographer's privacy.

[0004] In view of the problem of privacy disclosure, how to effectively remove or weaken the inherent properties of PRNU has attracted the attention of researchers. It has been proved that the PRNU of an image is difficult to be removed, and even if noise is added to the image or the effective bits of pixels are deleted from the image, it is also difficult to weaken the tracing ability of the image. Most of the previous image anonymization methods are directed to natural images, but as for biometric images, the existing methods greatly reduce the effectiveness of anonymization or have a large time complexity; moreover, the image itself contains various noises, among which the photo-response non-uniformity noise is used for image source device identification, and the PRNU noise must be extracted first. During the extraction process of PRNU noise, there is interference from scene details and other noises, and the existing methods cannot effectively remove these interferences to improve the accuracy of source device identification. Moreover, during the process of suppressing PRNU noise, the details of the image are easily affected, and the existing methods cannot ensure that the quality of the feature image does not decrease under high-frequency and low-frequency processing, minimize the impact on the image details, and achieve the best anonymization effect. SUMMARY

[0005] In order to solve the above problems, the present disclosure provides a photo-response non-uniformity noise anonymization method and system based on DCT and Wiener filtering. The PRNU anonymization algorithm based on discrete cosine transform (DCT) and Wiener filtering separately processes the high frequency of the image, ensures the effective removal of PRNU, and ensures the high quality of the image.

[0006] According to some embodiments, the present disclosure adopts the following technical solutions:

[0007] The photo-response non-uniformity noise anonymization method based on DCT and Wiener filtering comprises:

[0008] Step one: obtaining an image to be anonymized, and blocking the image;

[0009] Step two: performing discrete cosine transform on each block of the blocked image, separating the high frequency and low frequency regions of each block of the blocked image, extracting the high frequency component of each block of the blocked image and retaining the low frequency component, performing multiple Wiener filtering on the high frequency component, merging the high frequency component after multiple Wiener filtering with the retained low frequency component, and restoring the whole to a preliminary anonymous blocked image after merging using inverse discrete cosine transform.

[0010] Step three: determining whether each block of the blocked image has been restored to a preliminary anonymous blocked image, and if so, merging all the preliminary anonymous blocked images, and obtaining a final anonymous image by using multiple Wiener filtering again after merging.

[0011] According to some embodiments, the present disclosure adopts the technical scheme as follows:

[0012] A light response non-uniformity noise anonymization system based on DCT and Wiener filtering, comprising:

[0013] An image acquisition module for obtaining an image to be anonymized, and blocking the image;

[0014] A preliminary anonymous image acquisition module for performing discrete cosine transform on each block of the blocked image, separating the high frequency and low frequency regions of each block of the blocked image, extracting the high frequency component of each block of the blocked image and retaining the low frequency component, performing multiple Wiener filtering on the high frequency component, merging the high frequency component after multiple Wiener filtering with the retained low frequency component, and restoring the whole to a preliminary anonymous blocked image after merging using inverse discrete cosine transform.

[0015] A final anonymous image acquisition module for determining whether each block of the blocked image has been restored to a preliminary anonymous blocked image, and if so, merging all the preliminary anonymous blocked images, and obtaining a final anonymous image by using multiple Wiener filtering again after merging.

[0016] According to some embodiments, the present disclosure adopts the technical scheme as follows:

[0017] A computer readable storage medium, wherein a plurality of instructions are stored, the instructions being adapted to be loaded and executed by a processor of a terminal device, and the instructions being based on the light response non-uniformity noise anonymization method based on DCT and Wiener filtering.

[0018] According to some embodiments, the present disclosure adopts the technical scheme as follows:

[0019] A terminal device comprises a processor and a computer readable storage medium, the processor is used to implement instructions; the computer readable storage medium is used to store a plurality of instructions, the instructions are suitable for being loaded and executed by the processor and the DCT and Wiener filter based PRNU anonymity method.

[0020] Compared with the prior art, the beneficial effects of the present disclosure are:

[0021] The present disclosure proposes a PRNU anonymity algorithm based on discrete cosine transform (DCT) and Wiener filtering to solve the privacy leakage problem occurring in image authentication, uses block-based discrete cosine transform, processes the high frequency of the image alone, ensures the high quality and details of the image under the premise of effectively removing PRNU.

[0022] The present disclosure successfully suppresses the connection between a given image and its shooting mobile phone under the condition of smaller algorithm time complexity, and ensures that the biometric feature utility of the biometric feature image is not affected. BRIEF DESCRIPTION OF DRAWINGS

[0023] The accompanying drawings, which form a part of this disclosure, are intended to provide further understanding of the present disclosure, and the schematic embodiments of the present disclosure and the description thereof are used to explain the present disclosure, and do not constitute improper limitation on the present disclosure.

[0024] Figure 1 The anonymity method provided for the embodiments of the present disclosure is shown in the overall flowchart;

[0025] Figure 2 The final pre- and post-anonymity comparison result graph of one device in the MICHE-I data set provided for the embodiments of the present disclosure is shown in the figure; wherein, Figure 2 (a) in the figure is the pre-anonymity image, Figure 2 (b) in the figure is the post-anonymity image;

[0026] Figure 3 The final pre- and post-anonymity comparison result graph of one device in the UBIRIS data set provided for the embodiments of the present disclosure is shown in the figure; wherein, Figure 3 (a) in the figure is the pre-anonymity image, Figure 3 (b) in the figure is the post-anonymity image. DETAILED DESCRIPTION

[0027] The present disclosure will be further described below in combination with the drawings and embodiments.

[0028] It should be pointed out that the following detailed description is exemplary and is intended to provide further description of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as generally understood by those skilled in the art to which the present disclosure belongs.

[0029] It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments in accordance with the present disclosure. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, steps, operations, elements, components, and / or groups thereof, but do not preclude the presence or addition of one or more other features, steps, operations, elements, components, and / or groups thereof.

[0030] Embodiment 1

[0031] In one embodiment of the present disclosure, a DCT and Wiener filter based light response non-uniformity noise anonymization method is provided, comprising:

[0032] Step one: obtaining an image to be anonymized, and dividing the image into blocks;

[0033] Step two: performing discrete cosine transform on each block of the divided image, separating the high frequency and low frequency regions of each block of the divided image, extracting the high frequency component of each block of the divided image and retaining the low frequency component, performing multiple Wiener filter processing on the high frequency component, merging the high frequency component after multiple Wiener filter processing and the retained low frequency component, and restoring the whole to a preliminary anonymized block image after merging;

[0034] Step three: determining whether each block of the divided image has been restored to a preliminary anonymized block image, if so, merging all the preliminary anonymized block images, and obtaining a final anonymized image by using multiple Wiener filter processing again.

[0035] As an embodiment, PRNU is widely present in the high frequency region of the image, and a small amount is distributed in the low frequency region, and the low frequency region mainly saves the detail data of the image. Therefore, by a series of processing on the high frequency and retaining the low frequency region data, the effect of PRNU can be suppressed while the image quality is ensured, thereby achieving the effect of anonymity. The present disclosure proposes a DCT and Wiener filter based light response non-uniformity noise anonymization algorithm, which first divides the obtained image into blocks, uses the block division technology to refine the processing of the high frequency region, and uses Wiener filter on the high frequency region.

[0036] Wherein, the formula of the algorithm is as follows:

[0037] I' = 2F (φ' (2F (Φ (I) H ) + Φ (I) L ))

[0038] Wherein, I represents a block image of the original image after division, I' represents the image after anonymity, F represents Wiener filter, H represents high frequency component, L represents low frequency component, Φ represents discrete cosine transform, and φ' represents inverse discrete cosine transform.

[0039] As an embodiment, in step one, the process of obtaining the image to be anonymized and dividing the image into blocks includes:

[0040] Considering that processing the entire image may cause insufficient data processing, the algorithm proposed by the present disclosure first processes the image by dividing it into blocks. The size of the blocks has a significant impact on the efficiency of the anonymization algorithm: if the blocks are too large, the desired anonymization effect cannot be achieved, and if the blocks are too small, the visual quality (PSNR) of the processed image is low. After repeated testing, the present disclosure can maintain a high image quality and a low correlation when divided into 8x8 blocks, so 8x8 blocks are selected to maintain high image quality.

[0041] As an embodiment, in step two, the discrete cosine transform is applied to each block of the divided image, and the calculation formula is as follows:

[0042]

[0043] wherein,

[0044]

[0045]

[0046] A represents the input image, B represents the output result, M and N represent the two dimensions of the image, and p and q represent the dimensions of the output result.

[0047] The discrete cosine transform is applied to each block of the divided image, and the discrete cosine transform is symmetrical. Since the DCT transform is lossless and symmetrical, the high-frequency information in the image after the discrete cosine transform is concentrated in one area, and the low-frequency information is concentrated in one area. There is PRNU and image approximation information in the high-frequency information, so the high-frequency components of the high-frequency information are extracted for Wiener filtering.

[0048] Specifically, after the transformation of the divided image, the high-frequency information is mainly concentrated in the lower right corner of the transformation matrix, and the low-frequency information is mainly concentrated in the upper left corner of the transformation matrix. There is a large amount of PRNU and a small amount of image approximation information in the high-frequency information. In order to suppress PRNU as much as possible while achieving good image quality, the present disclosure extracts high-frequency information and low-frequency information respectively.

[0049] The present disclosure tests the median filter and the mean filter respectively, and the results show that both cannot achieve the anonymization effect and lose a large amount of image detail information. Considering that the Wiener filter is the optimal filter based on the minimum mean square error, and the test results are good, the present disclosure selects the Wiener filter.

[0050] In order to preserve the image detail information in high frequency, the present disclosure carries out Wiener filtering processing on the high frequency information of each block. Through repeated tests, it is found that when the Wiener filtering is used once, the high frequency information of the image cannot be well suppressed, and when the Wiener filtering is used three times, the visual quality of the image is reduced, so the present disclosure selects to use the Wiener filtering twice.

[0051] wherein the process of carrying out multiple times of Wiener filtering processing on the high frequency component is carrying out twice of Wiener filtering processing on the high frequency component. The filtering formula is:

[0052]

[0053] wherein b(n1, n2) represents the result after processing, v 2 represents the average of all local estimation variances, μ and σ 2 respectively represent the local mean and variance around each pixel. The mean calculation formula of the local neighborhood η of each pixel N x M in the image a is as follows:

[0054]

[0055] The variance calculation formula is as follows:

[0056]

[0057] Through repeated tests, neither too large nor too small filter window in each dimension can achieve the optimal effect, and finally the present disclosure selects the optimal 8 x 8 filter window.

[0058] As an embodiment, in step two, the step of extracting the high frequency component of each block of the partitioned image and preserving the low frequency component is to define a parameter to adjust the high frequency component, calculate the minimum value of the height and width of the partitioned image, multiply the minimum value by the parameter, and the product is used as the threshold value of the high frequency component and the low frequency component to extract the high frequency component.

[0059] Specifically, a parameter a is defined for adjusting the selection of the high frequency region, and the value range is [0, 1], and the present disclosure sets the parameter a to 0.9. After calculating the minimum value of the height (h) and width (w) of the test image and multiplying the parameter a, another parameter coff is obtained.

[0060] wherein coff = a x min(h, w)

[0061] wherein a is the defined parameter, h is the height of the test image, and w is the width of the test image.

[0062] The parameter coff is equivalent to the threshold value of the high frequency region and the low frequency region, which is used for high frequency suppression of each block after partitioning.

[0063] As an embodiment, the high frequency components processed by the multiple Wiener filtering are combined with the reserved low frequency components, and the process of reducing the preliminary anonymous block image as a whole by using the inverse discrete cosine transform (IDCT) after the combination of the processed high frequency information and the low frequency information is adding the processed high frequency information and the low frequency information, and reducing the preliminary anonymous image by using the inverse discrete cosine transform (IDCT).

[0064] wherein the inverse discrete cosine transform (IDCT)

[0065]

[0066] wherein α p , α q is consistent with the discrete cosine transform.

[0067] It is judged whether each block of the block image has been reduced to the preliminary anonymous block image, if yes, all the preliminary anonymous block images are combined, and the final anonymous image is obtained by using the multiple Wiener filtering as a whole after the combination.

[0068] If not, the above-mentioned processing is continuously performed on each block of the block image until each block of the block image forms the preliminary anonymous image.

[0069] However, after the combination, there is still a small amount of PRNU in the whole image, and the Wiener filtering processing is performed on the preliminary anonymous image at the end of the algorithm. It is found through tests that there is still a small amount of PRNU after one Wiener filtering, and the visual quality of the image is reduced after three times of filtering. Therefore, the present disclosure performs two times of Wiener filtering processing to obtain the final anonymous block image.

[0070] As an embodiment, the technical solution of the present disclosure is used to respectively anonymize one device in the MICHE-I dataset and one device in the UBIRIS dataset, wherein, Figure 2 (a), (b) and Figure 3 (a), (b) in the MICHE-I dataset and one device in the UBIRIS dataset are shown in the final anonymous image before and after the comparison chart, and it can be seen from the results that there is no significant difference in the image after the anonymization.

[0071] Embodiment 2

[0072] In an embodiment of the present disclosure, a light response non-uniformity noise anonymization system based on DCT and Wiener filtering is provided, comprising:

[0073] An image acquisition module is configured to acquire an image to be anonymized and block the image.

[0074] The preliminary anonymous image acquisition module is configured to perform discrete cosine transform on each block image, separate the high-frequency and low-frequency regions of each block image, extract the high-frequency component of each block image and retain the low-frequency component, perform multiple Wiener filtering on the high-frequency component, combine the high-frequency component after multiple Wiener filtering with the retained low-frequency component, and restore the combined whole into a preliminary anonymous block image by using inverse discrete cosine transform.

[0075] The final anonymous image acquisition module is configured to determine whether each block image has been restored into a preliminary anonymous block image, and if so, combine all the preliminary anonymous block images, and then obtain a final anonymous image by using multiple Wiener filtering on the combined whole.

[0076] The system described in Embodiment 2 specifically performs the specific implementation steps of the method described in the embodiment.

[0077] Embodiment 3

[0078] In an embodiment of the present disclosure, a computer readable storage medium is provided, in which a plurality of instructions are stored, the instructions being adapted to be loaded and executed by a processor of a terminal device to implement the steps of the light response non-uniformity noise anonymization method based on DCT and Wiener filtering described in Embodiment 1.

[0079] Embodiment 4

[0080] In an embodiment of the present disclosure, a terminal device is provided, which includes a processor and a computer readable storage medium, the processor being configured to implement instructions, and the computer readable storage medium being configured to store a plurality of instructions, the instructions being adapted to be loaded and executed by the processor to implement the steps of the light response non-uniformity noise anonymization method based on DCT and Wiener filtering described in Embodiment 1.

[0081] The present disclosure is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present disclosure. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the computer or other programmable data processing devices produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks

[0082] These computer program instructions can also be loaded into a computer or other programmable data processing devices, so that a series of operational steps are performed on the computer or other programmable data processing devices to generate a computer implemented process, so that the instructions executed on the computer or other programmable data processing devices provide a process for implementing the functions specified in the flowchart Figure 1 one flow or multiple flows and / or the functions specified in the block Figure 1 one flow or multiple flows and / or the functions specified in the block

[0083] The above describes the specific embodiments of the present disclosure in conjunction with the drawings, but is not a limitation on the protection scope of the present disclosure. Those skilled in the art should understand that various modifications or changes made by those skilled in the art on the basis of the technical solutions of the present disclosure without creative labor are still within the protection scope of the present disclosure.

Claims

1. A light response non-uniformity noise anonymization method based on DCT and Wiener filtering, characterized in that, The method comprises the steps of: Step 1: obtaining an image to be anonymized and dividing the image into blocks; Step 2: performing discrete cosine transform on each block of the divided image to separate high-frequency and low-frequency regions of each block of the divided image, extracting high-frequency components of each block of the divided image and retaining low-frequency components, performing multiple Wiener filtering on the high-frequency components, merging the high-frequency components after multiple Wiener filtering with the retained low-frequency components, and restoring the merged image as a preliminary anonymous block image by using inverse discrete cosine transform; The process of performing multiple Wiener filtering on the high-frequency components is to perform twice Wiener filtering on the high-frequency components; Step 3: determining whether each block of the divided image has been restored as a preliminary anonymous block image, and if so, merging all the preliminary anonymous block images and then obtaining a final anonymous image by using multiple Wiener filtering on the merged image; The process of merging all the preliminary anonymous block images and then obtaining a final anonymous image by using multiple Wiener filtering on the merged image is to obtain a final anonymous image by using twice Wiener filtering on the merged preliminary anonymous block image.

2. The DCT and Wiener filter based optical response non-uniformity noise anonymization method of claim 1, wherein, The image is divided into 8 blocks to maintain high image quality after repeated testing. 8 3. The DCT and Wiener filter based optical response non-uniformity noise anonymization method of claim 1, wherein, The step of extracting high-frequency components of each block of the divided image and retaining low-frequency components is to define a parameter to adjust the high-frequency components, calculate the minimum value of the height and width of the divided image, multiply the minimum value by the parameter, and use the product as a threshold value for separating high-frequency components and low-frequency components.

4. The DCT and Wiener filter based optical response non-uniformity noise anonymization method of claim 1, wherein, The discrete cosine transform is applied to each block of the divided image, and the discrete cosine transform is symmetrical.

5. The DCT and Wiener filter based optical response non-uniformity noise anonymization method of claim 4, wherein, High-frequency information is concentrated in one region and low-frequency information is concentrated in one region in the image after the discrete cosine transform, and PRNU and image approximate information exist in the high-frequency information, so the high-frequency components of the high-frequency information are extracted for separate Wiener filtering.

6. A light response non-uniformity noise anonymization system based on DCT and Wiener filtering, characterized in that, The method comprises the steps of: An image acquisition module is configured to obtain an image to be anonymized and divide the image into blocks; A preliminary anonymous image acquisition module is configured to perform discrete cosine transform on each block of the divided image to separate high-frequency and low-frequency regions of each block of the divided image, extract high-frequency components of each block of the divided image and retain low-frequency components, perform multiple Wiener filtering on the high-frequency components, merge the high-frequency components after multiple Wiener filtering with the retained low-frequency components, and restore the merged image as a preliminary anonymous block image by using inverse discrete cosine transform; The process of performing multiple Wiener filtering on the high-frequency components is to perform twice Wiener filtering on the high-frequency components; A final anonymous image acquisition module is configured to determine whether each block of the divided image has been restored as a preliminary anonymous block image, and if so, merge all the preliminary anonymous block images and then obtain a final anonymous image by using multiple Wiener filtering on the merged image; The process of merging all the preliminary anonymous block images and then obtaining a final anonymous image by using multiple Wiener filtering on the merged image is to obtain a final anonymous image by using twice Wiener filtering on the merged preliminary anonymous block image.

7. A computer readable storage medium characterized in that, A computer readable storage medium having stored therein a plurality of instructions, the instructions being adapted to be loaded and executed by a processor of a terminal device to implement the DCT and Wiener filter based light response non-uniformity noise anonymization method of any one of claims 1-5.

8. A terminal device, comprising: A computer system comprising a processor and a computer readable storage medium, the processor being configured to implement instructions; and the computer readable storage medium being configured to store a plurality of instructions, the instructions being adapted to be loaded and executed by the processor to implement the DCT and Wiener filter based light response non-uniformity noise anonymization method of any one of claims 1-5.

Images