CN-121998811-A - Document image watermark adding and falsification detection method based on Chebyshev and Arnold

CN121998811ACN 121998811 ACN121998811 ACN 121998811ACN-121998811-A

Abstract

The invention discloses a document image watermarking and tampering detection method based on Chebyshev and Arnold, wherein the watermarking method comprises the steps of converting a digital paper file into an image file, carrying out Arnold transformation on the image to obtain a scrambled image xi, converting the image xi into a Ycbcr color space, extracting a Y channel, carrying out two-dimensional Chebyshev transformation on the Y channel, reserving coefficients larger than a set threshold mu, setting the rest coefficients to zero, carrying out Chebyshev inverse transformation on the processed Y channel, replacing the Y channel into the Y channel in the Ycbcr space, converting the Y channel back into the image, carrying out Arnold inverse transformation, carrying out Chebyshev inverse transformation on a part with the coefficients smaller than the set threshold mu, and forming encrypted global summary information by extracting noise level characteristics. The invention hardly increases the image storage cost after watermark embedding, and can even obtain the storage smaller than the original file size under partial conditions, thereby effectively reducing the cost of storage and transmission.

Inventors

WANG SU
NIU KUN
LI YONG
TIAN LINGLING

Assignees

安徽产业互联数据智能创新中心有限公司

Dates

Publication Date: 20260508
Application Date: 20260107

Claims (10)

1. A document image watermarking method based on Chebyshev and Arnold, characterized in that the method comprises the steps of: Converting the digital paper file into an image file; arnold transformation is carried out on the image to obtain a scrambled image xi; Converting the image xi into a Ycbcr color space, and extracting a Y channel; Performing a two-dimensional Chebyshev transform on the Y channel; Reserving coefficients larger than a set threshold mu, and setting the rest coefficients to zero; and performing a Chebyshev inverse transform on the processed Y channel, converting it back to an image instead of the Y channel in Ycbcr space, and performing an Arnold inverse transform; The portion of the coefficient smaller than the set threshold μ is subjected to Chebyshev inverse transformation, and by extracting noise level features, encrypted global digest information is formed.
2. Document image watermarking method based on Chebyshev and Arnold according to claim 1, characterized in that in said step of Arnold transforming the input image, it comprises in particular: applying a plurality of sets of Arnold transforms with perturbation parameters (a_i, b_i) to the input image, wherein the perturbation parameters (a_i, b_i) are randomly generated; the disturbance parameters (a_i, b_i) are stored after being encrypted.
3. The document image watermarking method based on Chebyshev and Arnold as claimed in claim 1, wherein in the step of extracting noise level features, it specifically comprises: the result of performing the inverse Chebyshev transform on the portion of the coefficient smaller than the set threshold μ is recorded as ; Based on the formula Performing feature extraction operation, wherein K is an average convolution kernel and D is an expansion convolution kernel; and taking the noise maximum value obtained after the Toront iterative operation as the image global summary information, encrypting and storing.
4. The document image watermarking method based on Chebyshev and Arnold according to claim 2, wherein the encryption mode of the global summary information and the disturbance parameters is asymmetric encryption and is stored in a signature layer of the digital paper OFD format.
5. A document image falsification detection method based on Chebyshev and Arnold, characterized in that the detection method comprises the following steps: acquiring image data to be detected processed by the document image watermarking method based on Chebyshev and Arnold according to any one of claims 1 to 4; Arnold transformation is carried out on the image to be detected, and a scrambled image xi is obtained; Converting the image xi into a Ycbcr color space, and extracting a Y channel; Performing two-dimensional Chebyshev transformation on the Y channel, performing inverse Chebyshev transformation on a part with coefficients smaller than a set threshold mu, and extracting noise level characteristics; decrypting the image data to be detected to obtain global abstract information, and comparing the global abstract information with the extracted noise level characteristics; And if the local noise level exceeds the global summary information, judging that the corresponding position is tampered.
6. The document image falsification detection method based on Chebyshev and Arnold as claimed in claim 5, wherein the step of performing Arnold transformation on the image to be detected comprises: Decrypting the signature layer in the image data to be detected to obtain disturbance parameters (a_i, b_i); Arnold transformation is performed on the pattern to be detected based on the disturbance parameters (a_i, b_i).
7. A document image falsification detection system based on Chebyshev and Arnold, the system comprising: The watermark adding module is used for adding fragile watermark to the graphic file converted from the digital paper file, and concretely comprises the steps of performing Arnold transformation on the image to obtain a scrambled image xi, converting the image xi into a Ycbcr color space, extracting a Y channel, performing two-dimensional Chebyshev transformation on the Y channel, reserving coefficients larger than a set threshold mu, setting the rest coefficients to zero, performing Chebyshev inverse transformation on the processed Y channel, replacing the Y channel with the Y channel in the Ycbcr space, converting the Y channel back to the image, performing Arnold inverse transformation, performing Chebyshev inverse transformation on the part with the coefficient smaller than the set threshold mu, and forming encrypted global summary information by extracting noise level characteristics; The watermark extraction module is used for extracting noise level characteristics in the graph to be detected; the watermark detection module is used for decrypting the image data to be detected to obtain global abstract information, comparing the global abstract information with the extracted noise level characteristics, and judging that the corresponding position is tampered if the local noise level exceeds the global abstract information.
8. The document image falsification detection system based on Chebyshev and Arnold as claimed in claim 7, wherein the watermark extraction module extracts noise level features, specifically comprising: Arnold transformation is carried out on the image to be detected, and a scrambled image xi is obtained; Converting the image xi into a Ycbcr color space, and extracting a Y channel; A two-dimensional Chebyshev transform is performed on the Y channel, and a portion of the coefficient smaller than the set threshold μ is subjected to the Chebyshev inverse transform, and noise level features are extracted.
9. The document image falsification detection system based on Chebyshev and Arnold as claimed in claim 7, wherein in said step of extracting noise level features, specifically comprising: the result of performing the inverse Chebyshev transform on the portion of the coefficient smaller than the set threshold μ is recorded as ; And based on the formula Performing feature extraction operation, wherein K is an average convolution kernel and D is an expansion convolution kernel; and taking the noise maximum value obtained after the Toront iterative operation as the image global summary information, encrypting and storing.
10. The Chebyshev and Arnold based document image falsification detection system of claim 8, wherein: applying Arnold transformation with disturbance parameters (a_i, b_i) to an input image, wherein the disturbance parameters (a_i, b_i) are randomly generated; encrypting the disturbance parameters (a_i, b_i) and storing the encrypted disturbance parameters in a signature layer of the digital paper OFD format; the method specifically comprises the steps of decrypting a signature layer in image data to be detected to obtain disturbance parameters (a_i, b_i), and performing Arnold transformation on the image to be detected based on the disturbance parameters (a_i, b_i).

Description

Document image watermark adding and falsification detection method based on Chebyshev and Arnold Technical Field The invention belongs to the field of image falsification identification, and particularly relates to a document image watermark adding and falsification detection method based on Chebyshev and Arnold. Background The digital paper is a document format with data integrity verification of an OFD format file based on a pure image, combines the page-by-page hash verification function of a traditional digital signature, but cannot be accurate to a specific position of each page on tamper detection fine granularity. At present, image tampering detection is mainly divided into two main methods, namely a passive detection method represented by deep learning and an active detection method represented by fragile watermark. The two methods can meet the tamper detection requirements in different scenes to a certain extent, but still have obvious defects. The basic principle of the passive detection method represented by deep learning is to utilize the memory capacity of the model to the local tampering features of the image and to identify potential tampering by relying on the generalization capacity of the classification model. However, such a method of operation is highly dependent on a large-scale, high-quality training data set, and once the training data is different from the actual application scene, the detection accuracy is significantly reduced. In addition, in the case where the same image is tampered and modified and then detected again, it is difficult to effectively determine whether or not there is a new tampering behavior by passive detection. The active detection method represented by fragile watermark requires embedding local texture which is easy to be damaged, such as residual energy distribution of original image and low-frequency projection, local variance consistency and the like, in the image generation or use stage by an authorized person in advance. If the image is tampered with, these textures will change, thereby assisting in detecting the tampered position. However, the method has the problems that firstly, the semi-fragile watermark is difficult to strictly limit the detection authority in practical application, and only an authorized person embedded with the watermark cannot be guaranteed to effectively detect the watermark, and secondly, once the watermark embedding and extracting mechanism is disclosed, an attacker can pertinently bypass or destroy the watermark characteristics, so that the detection result is invalid. In summary, the existing image tampering detection method has defects in terms of robustness, controllability and safety, and is difficult to meet the requirements of high reliability and high safety in practical application. Disclosure of Invention Aiming at the problems, the technical scheme adopted by the invention is that the document image watermarking method based on Chebyshev and Arnold comprises the following steps: Converting the digital paper file into an image file; arnold transformation is carried out on the image to obtain a scrambled image xi; Converting the image xi into a Ycbcr color space, and extracting a Y channel; Performing a two-dimensional Chebyshev transform on the Y channel; Reserving coefficients larger than a set threshold mu, and setting the rest coefficients to zero; and performing a Chebyshev inverse transform on the processed Y channel, converting it back to an image instead of the Y channel in Ycbcr space, and performing an Arnold inverse transform; The portion of the coefficient smaller than the set threshold μ is subjected to Chebyshev inverse transformation, and by extracting noise level features, encrypted global digest information is formed. Optionally, in the step of performing Arnold transformation on the input image, the method specifically includes: applying a plurality of sets of Arnold transforms with perturbation parameters (a_i, b_i) to the input image, wherein the perturbation parameters (a_i, b_i) are randomly generated; the disturbance parameters (a_i, b_i) are stored after being encrypted. Optionally, in the step of extracting the noise level feature, the method specifically includes: the result of performing the inverse Chebyshev transform on the portion of the coefficient smaller than the set threshold μ is recorded as ; Based on the formulaPerforming feature extraction operation, wherein K is an average convolution kernel and D is an expansion convolution kernel; and taking the noise maximum value obtained after the Toront iterative operation as the image global summary information, encrypting and storing. Optionally, the encryption mode of the global summary information and the disturbance parameters is asymmetric encryption, and the signature layer of the digital paper OFD format is stored. A document image falsification detection method based on Chebyshev and Arnold, the detection method comprising the steps