41 results about "Embedding distortion" patented technology

The invention discloses a video steganography method based on a motion vector, which comprises the following steps: the type of each frame of a video sequence is judged; if the t-th frame is the key I frame, the motion vector collection of the frame is set up to null; if the t-th frame is the predictive P frame, motion estimation is performed to the frame to obtain the motion vector collection, and preset distortion function is used for performing embedded distortion definition to the motion vector collection; after the types of all the frames of the video sequence are judged and after the embedded distortion definition is performed to the motion vectors of all the P frames, syndrome-trellis codes (STC) are used for embedding secret information by taking all the motion vectors of the video sequence as carriers so as to obtain motion vector collection subjected to modification, and then compressed coding is performed to each frame of the video sequence in sequence to obtain the coding bit stream of the video sequence. Through the method disclosed by the invention, the anti-detection performance of embedding secret information is improved, and low bit rate expansion of the coding bit steam and the reconstruction of excellent visual quality of the video are ensured.

The invention discloses a color image reversible digital watermarking method based on a three-dimensional prediction error histogram. The method comprises steps: a watermark is embedded and extracted; through changing a prediction error group mapping relationship in the three-dimensional prediction error histogram, mapping with large embedding distortion is abandoned, mapping is divided into mapping whose embedding distortion is no more than 2 and mapping whose embedding distortion is no more than 1, and a reversible watermark is embedded in a new mapping relationship rebuilding method. Distortion of the image can be avoided to a certain degree, and the PSNR of the image is improved. The method can be used for hiding of secret information; and through applying a double-layer embedding scheme based on a diamond prediction model and building the prediction error group mapping relationship different from the previous one, less image distortion under the same embedding capacity can be realized.

A lossless, reversible data embedding technique uses a generalization of least-significant-bit modification. The portions of the signal that are susceptible to embedding distortion are compressed and sent as part of the embedded payload. A prediction-based conditional entropy coder uses unaltered portions of the host signal to improve lossless data capacity.

The invention discloses a reversible watermarking algorithm based on PVO (Pixel Value Ordering) and self-adaptive block segmentation and belongs to the field of multimedia information security. According to the algorithm, an image block is roughly segmented into two categories according to the complexity of local texture characteristics: a smooth block and a texture block; the smooth block is then subdivided into two sub-categories according to the correlation degree of a pixel in blocks: blocks with low smoothness and blocks with high smoothness. The blocks with low smoothness are ordered according to the sequence of pixel values from low to high to obtain psigma (1), psigma(2),to psigma (n); psigma(n-1) and psigma (n) are predicted by using psigma (n-2) or psigma(2) and psigma (1) are predicted by using psigma (3); one block with low smoothness can carry watermarking information of four bits at most. The blocks with high smoothness are segmented into sub-blocks which only contain three pixels respectively, that is (p1, p2 and p3), p1, p2 and p3 are ordered to obtain psigma (1), psigma (2) and psigma (3). According to the reversible watermarking algorithm based on PVO and self-adaptive block segmentation, the purpose of modifying psigma (1) and psigma (2) is achieved by only modifying psigma (3), the embedding distortion is greatly reduced.

The invention relates to a high-fidelity reversible watermark embedding method including a watermark embedding process, watermark extraction, and a raw image recovery process. During watermark embedding, pixels of a carrier image are divided into two pixel sets, A and B, which are not overlapped. The pixels belonging to the pixel set A are subjected to embedding first, then modified pixels in the pixel set A are used to predict pixels belonging to the pixel set B, and information embedding is conducted on the pixels in the pixel set B. For any to-be-predicted pixel p, n pixels surrounding the pixel p form a neighborhood of the pixel p, and after the values of all the pixels in the neighborhood are sorted, the maximum and the minimum in the neighborhood closest to the pixel p are used to predict the pixel p. As the pixels closest to the pixel p are used for predicting the pixel p, the prediction performance is greatly improved. The larger n is, the more precise the prediction can be, and the lower the capacity is accordingly. The smaller n is, the less precise the prediction is, and the higher the capacity is accordingly. With determined capacity, the method searches for the optimal embedding point with the least loss, thereby improving embedding distortion effectively.

The invention provides a reversible information hiding method based on self-adaptive expansion of a two-dimensional prediction error histogram, and the method comprises the steps: dividing an originalimage into a black pixel layer and a white pixel layer in a checkerboard mode, and sequentially carrying out the following steps on the black pixel layer and the white pixel layer; pairing the adjacent pixels in the same layer of pixels, predicting pixel values, and obtaining a two-dimensional prediction error histogram; searching an optimal reversible mapping strategy through an adaptive optimization method according to the two-dimensional prediction error histogram; and performing secret information embedding on the image according to the optimal reversible mapping strategy. According to the method, the optimal reversible mapping strategy is found out in a self-adaptive optimization mode, on the premise that the embedding capacity is guaranteed, embedding distortion is reduced, and theembedding performance is improved.

The invention provides a reversible data hiding method based on pixel multi-scale prediction, and the method comprises the steps: calculating the neighborhood complexity of each pixel on a carrier image in an area, and carrying out the classification of all pixels according to the neighborhood complexity of the pixels; Performing multi-scale prediction on the classified pixel set with a certain range of neighborhood complexity to obtain a prediction value, and further obtaining a prediction error value and a corresponding prediction error histogram; Using the histogram expansion- translation embedding algorithmto expand and translate the pixel values corresponding to the prediction error at the peak of the prediction error histogram,, and embedding the secret data into the prediction errorof the pixel value so as to obtain an image embedded with the secret data; And embedding the auxiliary information into the image embedded with the secret data through an LSB steganography algorithmto obtain an encrypted image containing the secret data. According to the method, under the condition that a certain embedding capacity is ensured, the embedding distortion of the carrier image is effectively reduced by fully utilizing the information redundancy among the pixels.

The invention discloses a JPEG image reversible information hiding method. The method includes: carrying out entropy decoding on the original JPEG image, converting DCT image, and obtaining a quantized DCT coefficient and a quantized table; calculating an embedded distortion value of each DCT block; carrying out information embedding on the DCT block, wherein the embedding sequence is that embedding distortion values are embedded from small to large; and carrying out entropy coding on the modified DCT coefficient, and obtaining a secret-carrying JPEG image. According to the method, the DCT blocks are selected and reversible information hiding is carried out according to distortion value sorting, so that the visual distortion of the secret-carrying JPEG image is effectively reduced.

The invention discloses an image steganography method and system for automatically learning distortion based on a GAN network structure, and the method comprises the steps: inputting an enhanced carrier image into a preset improved U-Net network, sequentially generating an initial pixel change probability graph and an initial secret-containing image, inputting the initial secret-containing image into a discrimination network, and obtaining a discrimination result, calculating the loss of the discrimination network according to the discrimination result, calculating a total loss function of the generative network according to the discrimination result and the current steganography capacity, optimizing the generative adversarial network by taking minimization of the loss function as a target, when the loss is reduced and kept stable, considering that the training is ended, extracting the generative network from the generative adversarial network after the training is ended, and inputting a to-be-transmitted original image into the generative network to obtain a pixel change probability, calculating embedded distortion corresponding to the pixel change probability, and coding the secret information and the to-be-transmitted original image by adopting a syndrome matrix coding technology according to the embedded distortion to obtain a steganographic image corresponding to the original image.

The invention discloses a high volume reversible watermarking method based on a self-adaptive prediction model. The method comprises a watermark embedding process and a watermark extraction and original image recovery process. The watermark embedding process comprises a design process of a prediction model, and a self-adaptive embedding strategy; and the watermark extraction and original image recovery process is the reverse process of the watermark embedding process. According to the invention, the characteristic of high correlation between neighboring pixels is fully utilized so as to obtain the prediction model with higher prediction performance, so that the high-embedding distortion generated when a prediction error is modified for watermark embedding in a conventional algorithm can be effectively reduced, and besides, through estimating the correlation degree of the four encircling pixels of each pixel to be embedded in a set, the pixels accounting for about three fourth of all the pixels can be enabled to carry 1-2 bit watermarks, thus the capacity is improved.

The invention discloses a high-capacity reversible watermark embedding and extracting method, which mainly includes the following steps: calculating prediction errors of pixel predicted values of all pixels in an image; constructing a statistic column diagram according to the prediction errors, dividing the pixels in the image into an embeddable pixel set embedded with watermark data, and a movable pixel set capable of moving horizontally; and embedding the watermark data into the pixels with the prediction errors lower than a set threshold value. During embedding, one bit or multiple bits of watermark information is embedded into a single pixel in a self-adaptive manner according to the texture complexity of the image, so that the single embedding rate can achieve more than 2BPP (bits per pixel), and the embedding capability is efficiently improved. In addition, even multiple bits of watermark information are embedded into the single pixel, the used prediction error are still the original prediction errors, therefore, the embedding distortion is efficiently lowered, namely, the higher image quality is ensured on the premise that the embedding capability is efficiently improved.

The invention discloses a high-volume reversible watermarking method based on remainder invariance. The method comprises a watermark embedding process and a watermark extraction and original image recovery process. A integer transformation provided by the invention removes additive terms and reduces embedding distortion; the integer transformation is composed of only a prediction process, and involves predicting each pixel in a block by using the mean value of the block, enabling each prediction error before a final prediction error in a sub-block to carry watermarking information and enabling a final pixel to be only predicted but carry no watermarking information so as to ensure the reversibility of the integer transformation; and by using the characteristic that the remainder obtained through the mean value of an image sub-block being divided by a given embedding parameter remains untransformed beforehand and afterwards, the integer transformation can help to recover original pixels and watermarking information without distortion. Therefore, by the provided method, improvements to a great extent are made in terms of both embedding capacity and recovery image quality.

The embodiment of the invention discloses a method and device for embedding and extracting a reversible watermarking as well as a method and device for recovering an image and relates to the technical field of image information safety. The method for embedding the reversible watermarking provided by the embodiment of the invention comprises the following steps: acquiring congregation classification information and watermarking information of a macro-block in an original image, wherein the congregation classification information comprises a first congregation, a second congregation and a third congregation; establishing a position chart related to an embedded quantity for each macro-block in the first and second congregations; and embedding the information in the position chart related to the embedded quantity and the watermarking information into the corresponding macro-blocks, according to the arrayed sequence of the macro-blocks and the congregation classification of the macro-blocks. By using the method provided by the embodiment of the invention, the embedding volume is increased and the embedding distortion is reduced.

The invention discloses a reversible information hiding method based on improved pixel local complexity calculation and multi-peak embedding. The method comprises the following steps: carrying out overflow / underflow preprocessing on an original image; dividing the image into a Blue sub-image and a Blank sub-image, dividing pixel positions of the sub-images, and calculating a prediction error of a target pixel by using a diamond predictor; calculating the pixel texture degree of the Blue / Blank sub-images to obtain the texture degree of an adjacent pixel of the target pixel, and then calculating the local complexity of the target pixel by using the texture degree of the adjacent pixel; according to a prediction error histogram of each sub-image, selecting three peak points and two zero points closest to the maximum peak point and the minimum peak point, then adaptively selecting the peak points according to the secret information length, and preferentially carrying out information embedding on pixels with low local complexity in each sub-image; carrying out secret information extraction and image restoration. According to the invention, the image embedding capacity is improved, and the embedding distortion is reduced.

The invention discloses an image steganography method based on Gibbs sampling. Firstly, array segmentation is carried out on a steganography image based on a field clustering idea, and Gibbs sampling and STC coding are carried out on segmented sub-arrays; and secondly, direction information (including horizontal, vertical and opposite angle information) of sub-array images is fully utilized, a weighted summation mode is adopted to obtain a steganography distortion function, and higher-order statistic characteristics of the image is maintained. A Gibbs sampling theory is utilized to realize information embedding, so that the embedding distortion is minimized, and the characteristic model of the steganography image is maintained; in addition, function parameters are optimized in a Fisher rule frame, so that the excessive optimization problem of an image characteristic set is effectively avoided. Shown by experiments, the distortion of the image model is reduced, the classification error of the steganography image is effectively improved, and the steganography safety is enhanced.

The invention discloses a high-capacity data protection method based on a GAN augmented image foreground object. By applying various GANs to information hiding, the high-capacity data protection method obtains an embedded distortion probability graph, calculates an unsuitable steganography area according to the embedded distortion probability graph, generates a foreground object in the unsuitablesteganography area through a fixed area object generation and information embedding module, and then hides the secret information in the generated graph. The high-capacity data protection method mainly comprises the following steps: (1) inputting an original image into a steganography probability graph generator; (2) calculating the steganography probability graph; (3) calculating the unsuitable steganography area; (4) adding a texture object into the original image; and (5) carrying out an information hiding algorithm to embed secret information. According to the high-capacity data protectionmethod, the unsuitable hidden area is calculated for the image in advance, so that the directional hiding capability of the unsuitable hidden area of the original image in information hiding is improved; and provided is a high-capacity data protection method which increases hiding capacity and has high security.

The invention discloses a video steganography method based on motion vector embedding distortion decomposition, which comprises the following steps of: performing motion estimation on a video sequence, calculating the joint embedding distortion of a motion vector in an original motion vector field of a P frame by combining a distortion function, and further calculating the joint modification probability of the corresponding motion vector; according to the joint modification probability, calculating an edge modification probability of a horizontal component and a conditional modification probability of a vertical component of the motion vector so as to embed a load into the horizontal component and the vertical component of the motion vector; and forming a motion vector field modified by the video sequence by the motion vector embedded with the load, and generating a video coding bit stream through coding to complete video steganography. Dynamic load distribution in motion vector components is realized and anti-detection performance of the video steganography method is improved.

The invention provides a self-adaptive reversible watermark method and system. The method comprises the following steps of: dividing a carrier image into a plurality of adjacent pixel pairs which arenot coincident; calculating corresponding difference values by utilizing a gradient prediction method and generating a two-dimensional difference value histogram; selecting an optimum path in the two-dimensional difference value histogram by adoption of a dynamic planning algorithm, and establishing a first corresponding relationship between pixels corresponding to a difference value pair on the optimum path and bit information carried by a to-be-embedded watermark; if the difference value corresponding to the pixel pair is not on the optimum path, modifying the pixel pair according to a preset first rule; if the difference value corresponding to the pixel is on the optimum path and a value of the bit information corresponding to the pixel pair is 1, modifying the pixel pair according to the preset first rule and the first corresponding relationship; and if the value of the bit information corresponding to the pixel pair is 0, not modifying the pixel pair. The method and system are high in data embedding capacity, low in embedding distortion and high in hidden data safety.

The invention discloses a reversible watermark implementation method of an audio signal. The method comprises steps that a value of a left channel signal point is predicted through utilizing a first prediction model to obtain a residual error; the residual error is utilized to distinguish different smooth regions of a right channel signal, and an optimal right channel signal slip region is selected through utilizing the optimization algorithm; a value of each point in the optimal right channel signal slip region is predicted through utilizing a second prediction module, a residual error is obtained, and the spread method is utilized for embedding and extraction recovery of the watermark information. The method is advantaged in that the audio embedding capacity is improved while audio embedding distortion is greatly reduced, an original audio can be restored losslessly after watermark extraction, and the method can be applied to authenticity and nondestructive authentication of audio signals.

The embodiment of the invention discloses a reversible watermarking method and device based on a human vision system. The method comprises the following steps: dividing an original image into a texture complex block, a common block and a smooth block according to a preset first threshold value, a second threshold value greater than the first threshold value and the local complexity of each matrixblock of the original image; calculating the prediction error of the maximum pixel value, the second maximum pixel value, the minimum pixel value and the second minimum pixel value of each common block and each smooth block by adopting a pixel value prediction error method for calculating the prediction error of the second maximum pixel value based on the relative position relation of the maximumpixel value and the second maximum pixel value; embedding watermark information into each common block by adopting an ipvo watermark embedding method according to the calculated prediction error; andembedding watermark information into each smooth block by adopting a corresponding relation between a preset prediction error value and an embedded watermark pixel value to obtain a watermark embeddedimage. According to the method and the device, prediction is more accurate, and reversible watermark processing of embedding distortion introduced by gray level translation can be effectively reduced.

The invention discloses a reversible information hiding method based on improved difference expansion. Embedding is carried out on the brightness residual QDCT coefficients of the I frame, every two adjacent QDCT coefficients in each 4*4 block form a pair, coefficient difference values are extracted, the coefficient difference value pairs of the to-be-embedded information are modified, and the purpose of secret information embedding is achieved while the difference values are expanded. Due to the application of an intra-frame prediction mode in H.264 video coding, modification of a residual coefficient causes intra-frame distortion drift, firstly, brightness residual coefficient blocks suitable for embedding are screened according to the intra-frame prediction process and the prediction mode, and the intra-frame distortion drift can be inhibited. Through the improved reversible algorithm, the amplitude of double expansion of the difference value during embedding can be controlled, and the modification amplitude of the coefficient is reduced; and by utilizing the characteristic that the medium-high frequency coefficient value is small after DCT coefficient quantization in video coding, the embedding distortion is reduced, and the higher secret-carrying video quality is obtained.

The invention discloses an image spatial domain adaptive watermark based on a JND model. A digital image watermark system mainly comprises two main steps of 1, determining a watermark embedding scheme; 2, the watermark embedding strength is determined, and under the condition of ensuring a certain embedding capacity, the embedding strength of the airspace watermark is adaptively determined by fully considering the relationship between the current pixel value and the adjacent pixels, so that the imperceptibility of the watermark is improved, and the embedding distortion of the image is effectively reduced. Compared with the prior art, the method considers the self-adaptive determination of the embedding intensity for different embedding areas instead of adopting the same embedding intensity, so that the imperceptibility of the watermark can be more effectively improved, and the embedding distortion is reduced. For the determination of the watermark embedding intensity, the method not only considers the current single pixel value, but also considers the relationship with the adjacent pixels. According to the invention, the embedding capacity is satisfied, the embedding strength is maximized, the embedding distortion is minimized, and the embedding performance is improved.

The invention relates to a carrier selection method capable of avoiding preprocessing, which comprises a preprocessing distortion calculation stage, an embedded distortion calculation stage and a distortion fusion and carrier selection stage, and is characterized in that a pre-classifier is trained to give the possibility that an image is preprocessed, the higher the possibility is, the larger thepreprocessing distortion is, and for the embedded distortion, the current distortion minimization framework is used for calculation. Finally, the two types of distortion are fused, and an image withminimum total distortion is selected as a carrier. According to the carrier selection method capable of avoiding preprocessing, the selection of the preprocessed image as the carrier is avoided by avoiding the preprocessed image, so that the multi-carrier steganography security is improved.

The invention discloses a high-capacity reversible watermarking method based on an adaptive prediction model; it includes a watermark embedding process, watermark extraction and original image restoration process; the watermark embedding process includes a prediction model design process and an adaptive embedding strategy; The process of watermark extraction and original image recovery is the inverse process of the watermark embedding process; the present invention obtains a prediction model with high prediction performance by making full use of the characteristics of high correlation between adjacent pixels, thereby effectively reducing the The high embedding distortion generated when watermark embedding is modified by the prediction error; in addition, by evaluating the correlation degree of the four surrounding pixels of each pixel in the set to be embedded, about three quarters of all pixels can carry 1‑ A 2-bit watermark, therefore, achieves an increase in capacity.

The invention discloses a reversible watermarking algorithm based on PVO (Pixel Value Ordering) and self-adaptive block segmentation and belongs to the field of multimedia information security. According to the algorithm, an image block is roughly segmented into two categories according to the complexity of local texture characteristics: a smooth block and a texture block; the smooth block is then subdivided into two sub-categories according to the correlation degree of a pixel in blocks: blocks with low smoothness and blocks with high smoothness. The blocks with low smoothness are ordered according to the sequence of pixel values from low to high to obtain psigma (1), psigma(2),to psigma (n); psigma(n-1) and psigma (n) are predicted by using psigma (n-2) or psigma(2) and psigma (1) are predicted by using psigma (3); one block with low smoothness can carry watermarking information of four bits at most. The blocks with high smoothness are segmented into sub-blocks which only contain three pixels respectively, that is (p1, p2 and p3), p1, p2 and p3 are ordered to obtain psigma (1), psigma (2) and psigma (3). According to the reversible watermarking algorithm based on PVO and self-adaptive block segmentation, the purpose of modifying psigma (1) and psigma (2) is achieved by only modifying psigma (3), the embedding distortion is greatly reduced.