48 results about "Dct coefficient" patented technology

The invention discloses a screen content video coding perception code rate control method and device based on DCT domain JND, and relates to the field of video coding, and the method comprises the steps: obtaining a screen content video, and obtaining a DCT coefficient of the video through discrete cosine transform; guiding frame-level and CTU-level target bit allocation by using the texture energy weight; obtaining a JND threshold value of the DCT domain according to a product based on the spatial contrast sensitivity threshold value, the brightness masking modulation factor and the contrast masking adjustment factor; and deducing a Lagrange multiplier combined with a JND regulation factor by using a JND threshold value to realize code rate control. According to the method, the JND model suitable for the screen content video is established in the DCT domain, bit distribution is guided through the texture energy weight, the new rate distortion model is established, the coding rate distortion performance can be improved, and the code rate control precision of the screen content video is improved.

The invention discloses a document image tampering detection method based on text aggregation and multi-frequency enhancement, and relates to the technical field of computer vision, image forensics and deep learning, and the method comprises the steps: obtaining a to-be-detected original RGB image, carrying out the multi-mode decomposition of the to-be-detected image, and obtaining a DCT coefficient graph, a corresponding quantization table, a high-frequency view and a low-frequency view; and inputting the original RGB image, the DCT coefficient graph, the corresponding quantization table, the high-frequency view and the low-frequency view into a document image tampering detection model for processing, and outputting a final detection result. The document image tampering detection model carries out feature dimension reduction and preliminary text aggregation through the vision-frequency fusion module, carries out coding and fusion through the multi-frequency feature extractor, generates comprehensive frequency features, carries out wavelet transform decoupling through the direction perception frequency decoupling enhancement module, and outputs tampered area masks based on the decoding prediction module. According to the invention, hidden tampering artifacts can be revealed more comprehensively.

The invention relates to the technical field of video noise reduction, in particular to an ultra-high-definition video noise reduction method. The method comprises the following steps: acquiring video data and noise variance; obtaining a gray threshold and a gradient threshold based on the gray value and the gradient value of each frame of image pixel point of the video data, obtaining the region of each frame of image, and then determining a growth criterion through the gray difference and the gradient difference to block the image; determining a gray feature weight and a texture feature weight based on the difference between the gray value and the gray limit value and the difference between the gradient value and the gradient limit value; improving the original DCT coefficient based on the feature weight and the texture feature weight to construct an exclusive dictionary, and denoising, decoding and reconstructing the image based on the exclusive dictionary; and synthesizing each frame of reconstructed image into a complete video. According to the invention, optimal noise reduction of the video image frame is realized and detail information of the video is reserved to the greatest extent.

The invention discloses a method for detecting tiny defects of a high-density integrated circuit packaging substrate. The method comprises the following steps: acquiring image data of the surface of the high-density integrated circuit packaging substrate; processing the acquired images in sequence by adopting a mean filtering method, morphological dilation and a texture operator; processing the image by adopting an image binarization processing technology to obtain a binarized image; dividing the binary image into a plurality of energy regions, and calculating an entropy difference through the divided energy regions; performing processing operation on the image data through a discrete cosine transform technology to obtain a DCT coefficient, dividing the DCT coefficient into a high-frequency component and a low-frequency component, and performing calculation to obtain a ratio of a low-frequency coefficient mean value to a high-frequency coefficient mean value; the feature extraction model extracts defect shape features based on the entropy difference and the ratio; comparing a similarity measurement index calculated according to the defect shape characteristics of the packaging substrate with a preset threshold value, and judging whether the packaging substrate has a shape defect or not; and precise detection of small defects of the high-density integrated circuit packaging substrate is realized.

A method for disparity estimation between digital images includes generating an interleaved image from two or more digital images with an offset between them, where the interleaved image is subdivided into a plurality of patches, computing discrete cosine transform (DCT) coefficients of each of the plurality of patches, computing, for each of the plurality of patches, a mean DCT descriptor from the DCT coefficients of each patch, and determining a disparity map from the mean DCT descriptor of each of the plurality of patches using a classifier. The disparity map is configured for real-time depth estimation from the two or more digital images.

The invention provides a residual DCT (Discrete Cosine Transformation) feature guided artifact learning module and an image tampering detection system, and the module comprises a DCT coefficient processing unit, a quantization table processing unit, a residual DCT calculation unit, a feature splicing unit and a neural convolution processing unit. The residual DCT calculation unit is used for performing repeated recompression on DCT coefficient matrix simulation JPEG compression to obtain DCT residual information and construct a residual DCT feature map on the basis of the DCT residual information, and the DCT coefficient processing unit is used for encoding the DCT coefficient matrix into DCT plane binary volume representation by using threshold truncation and One-Hot encoding, and then performing feature extraction and processing; the quantization table processing unit is used for expanding and processing a quantization table; the feature splicing unit is used for feature splicing; and the neural convolution processing unit is used for processing the spliced features to obtain final DCT features. According to the method, the positioning and identification capability of the tampered area in the double JPEG image with the same QF compressed twice is effectively improved.

The present application relates to the technical field of image tampering detection and positioning, and relates to an image tampering detection and positioning method based on multi-domain fusion and dynamic quantization perception. The method comprises the following steps: S1: obtaining RGB pixel data, Y channel DCT coefficients and a quantization table of a to-be-detected image to obtain a to-be-detected data set; S2: performing multi-dimensional feature extraction on the to-be-detected data set to obtain a multi-dimensional feature set; S3: processing the multi-dimensional feature set to obtain a fused multi-dimensional feature tensor; S4: constructing a detection and positioning model based on multi-domain fusion and dynamic quantization perception; S5: performing two-stage training on the constructed detection and positioning model using a labeled image tampering sample data set; and S6: inputting the processed multi-dimensional feature tensor into the trained model to output a pixel-level tampering positioning result, thereby completing the image tampering detection and positioning. The present application has the advantages of significantly improved detection and positioning accuracy.

The disclosure provides a feature point detection-based anti-screen camera robust watermark embedding, extraction method and system, relates to the technical field of digital media copyright protection, and when a carrier image is photographed and watermark is extracted, an image correction algorithm based on PCA and SURF is proposed, the generated distortion can be well corrected, and the robustness of the screen camera watermark is indirectly ensured. The image after the screen camera can be corrected to restore the image to the state in the screen, and the selection of the feature area before and after the correction is also performed to improve the embedding effect of the watermark. In order to improve the robustness of the screen camera watermark, in the embedding of the screen camera watermark, the carrier image is subjected to DCT transformation, and then experiments are designed to select the DCT coefficients of the embedded watermark. Thus, after the watermark is embedded, the robustness is high, and the invisibility is also high.

The invention discloses an image recompression method and device, electronic equipment and a storage medium, and relates to the technical field of image coding, and the method comprises the steps: extracting metadata and an original DCT coefficient from an original image file, and the original image file is in a JPEG format; and determining a target deep learning model based on the metadata, processing the original DCT coefficient according to the target deep learning model, and generating a feature bit stream and a prediction image. And based on the prediction image, performing residual calculation on the original image file and the feature bit stream to obtain a residual bit stream. And combining and packaging the predicted image and the residual bit stream to obtain an image recompression file. By calculating the residual error between the predicted image and the original image, the focus of information compression is transferred from total image information to extremely small error information, so that the image can be reconstructed losslessly and completely at a decoding end according to a small amount of data and differential information, and the storage and transmission cost is effectively reduced while the image information completeness is ensured.

The invention discloses a color correction method and system, and belongs to the technical field of image processing, and the method comprises the steps: extracting a chromaticity quantization DCT coefficient, an initial chromaticity quantization table and a standard color card calibration matrix of a compressed image; wherein the chroma quantization DCT coefficient comprises a DCT coefficient of a blue chroma component and a DCT coefficient of a red chroma component; generating a chromaticity correction matrix according to the DCT coefficient of the blue chromaticity component of the compressed image, the DCT coefficient of the red chromaticity component of the compressed image and the standard color card calibration matrix; correcting the frequency domain chromaticity of the compressed image according to the chromaticity correction matrix; generating a final chroma quantization DCT coefficient according to the initial chroma quantization table and the chroma correction matrix; outputting a compressed image code stream according to the final chromaticity quantization DCT coefficient and the entropy coding table; according to the method, chromaticity correction is directly carried out in the frequency domain of the compressed image, quantization table optimization and chromaticity statistical feature modeling are combined, the problems of calculation complexity and distortion caused by decompression processing are avoided, and the color fidelity of the compressed image is improved.

The invention provides a JPEG (Joint Photographic Experts Group) compression artifact learning module and an image tampering detection system, and the JPEG compression artifact learning module comprises a DCT (Discrete Cosine Transform) coefficient processing unit, a quantization table processing unit, a feature splicing unit and a neural convolution processing unit, the DCT coefficient processing unit is used for coding a DCT coefficient matrix into DCT plane binary volume representation by using threshold truncation and One-Hot coding, and then performing feature extraction and processing to obtain a first feature map; the quantization table processing unit is used for expanding and processing the quantization table to obtain a second feature map; the feature splicing unit is used for splicing the first feature map and the second feature map; and the neural convolution processing unit is used for performing channel compression, normalization and activation on the spliced feature vectors to obtain final DCT features. According to the method, the CNN can directly learn a local JPEG compression artifact mode, and the tampering detection efficiency and precision are improved.

A method of encoding a video stream in a video encoder is provided that includes computing an offset into a transform matrix based on a transform block size, wherein a size of the transform matrix is larger than the transform block size, and wherein the transform matrix is one selected from a group consisting of a DCT transform matrix and an IDCT transform matrix, and transforming a residual block to generate a DCT coefficient block, wherein the offset is used to select elements of rows and columns of a DCT submatrix of the transform block size from the transform matrix.

A method and apparatus for perceptual bitrate control of screen content video coding based on Discrete Cosine Transform (DCT) domain JND is disclosed, relating to the field of video coding. The method includes: acquiring screen content video; obtaining DCT coefficients of the video through Discrete Cosine Transform; using texture energy weights to guide frame-level and CTU-level target bit allocation; obtaining a JND threshold in the DCT domain based on the product of a spatial contrast sensitivity threshold, a luminance masking modulation factor, and a contrast masking adjustment factor; and deriving a Lagrange multiplier for the JND adjustment factor using the JND threshold to achieve bitrate control. This invention establishes a JND model suitable for screen content video in the DCT domain, guides bit allocation through texture energy weights, and establishes a new rate-distortion model, which can improve coding rate-distortion performance and enhance the bitrate control accuracy of screen content video.

In one embodiment, an image feature is extracted that includes a lower-upper PSNR calculated between an original image and an image reduced (then re-scaled to an original resolution), an HOG (Histogram of Oriented Gradients) feature calculated on an image block extracted from the original image, and a DCT coefficient calculated on an image block extracted from the original image, and the DCT coefficient is also calculated on the image block. For DCT coefficients, only a subset at a high frequency is used. These image blocks are aggregated and serially connected before being fed to a neural network to predict QP (Quantization Parameter) switching values. If the QP value of the current picture is greater than the predicted QP switching value, RPR (Reference Picture Resampling) is applied to the current inter picture.

The application relates to the technical field of digital media processing, in particular to a robust steganography method and device for an image with a high quality factor and a large size, wherein the method comprises the following steps: decoding a carrier image to obtain first spatial domain pixel values of the carrier image; performing pre-scaling processing on a spatial domain image corresponding to the carrier image based on channel characteristics of a target network platform and obtaining discrete cosine transform (DCT) coefficients; embedding target secret information into the DCT coefficients to obtain initial carrier-cipher DCT coefficients, adjusting unstable coefficients of the initial carrier-cipher DCT coefficients, and obtaining final carrier-cipher DCT coefficients; transforming the final carrier-cipher DCT coefficients into a spatial domain to obtain second spatial domain pixel values, and modifying the first spatial domain pixel values according to the second spatial domain pixel values until an air domain image corresponding to the modified first spatial domain pixel values satisfies a first preset scaling condition; and compressing the air domain image corresponding to the modified first spatial domain pixel values to generate a final carrier-cipher image satisfying a preset size.

Example techniques and devices for coding media data are described. An example device for encoding media data includes one or more memories configured to store the media data and one or more processors implemented in circuitry and coupled to the one or more memories. The one or more processors are configured to determine a hyperprior variable of a neural video coder based on the media data. The one or more processors are configured to apply a discrete cosine transform (DCT) to the hyperprior variable to generate DCT coefficients and encode the DCT coefficients.

The invention belongs to the technical field of data processing, and discloses a color JPEG reversible watermark hiding method and system based on adaptive STC, and the method comprises the steps: obtaining a color JPEG image, decoding the color JPEG image, converting the color JPEG image into a Y color space, and extracting the DCT coefficients of the Y channel and the Y channel; according to the statistical characteristics of the DCT coefficients of all the channels, the embedding adaptability weight is calculated, and the embedding capacity is adaptively allocated to the three channels in combination with a preset channel weighting adjustment factor; a candidate set is selected according to the sequence of the three channels and the distributed embedding capacity, and zero value coefficients in the candidate set form a carrier sequence; the watermark to be hidden is embedded into the carrier sequence through STC coding, the secret-carrying JPEG image is generated, and high-capacity, low-distortion and completely reversible watermark hiding facing the color JPEG image is achieved.

The invention relates to the technical field of image coding, in particular to a variable-granularity self-adaptive coding method for a smoke and dust disaster environment, which comprises the following steps: performing dust infrared transmissivity dynamic compensation based on the obtained current dust concentration and visibility of a smoke and dust disaster environment scene, and calculating a transmissivity compensation coefficient; synchronously acquiring a visible light image and an SWIR image of the smoke and dust disaster environment scene, and performing gain compensation on the SWIR image based on the transmissivity compensation coefficient to obtain a compensated SWIR image; performing feature extraction and fusion on the visible light image and the compensated SWIR image to obtain fusion features, and restoring the fusion features into a fusion image; selecting the size of a DCT block in combination with the local texture complexity, and performing DCT on the fused image to obtain a group of DCT coefficients; and applying quantization parameter threshold constraint to the abnormal high-frequency DCT coefficient, and carrying out variable granularity DCT quantization on the fused image to obtain entropy coding.

The application provides a point-sol gel fabric manufacturing method, equipment and medium, relates to the point-sol gel fabric manufacturing technical field, and comprises the following steps: dividing a target image into a plurality of sub-images; performing DCT transformation on each sub-image to obtain a corresponding DCT coefficient matrix of each sub-image; clustering all DCT coefficient matrices according to the characteristics of each DCT coefficient matrix; processing each DCT coefficient matrix in QA to obtain a processed DCT coefficient matrix cluster WA corresponding to QA; mapping the energy center corresponding to each DCT coefficient matrix in WA to the target image to obtain the centroid of each DCT coefficient matrix in QA; obtaining the centroid of a complete glue point composed of a defective glue point; determining whether the point-sol gel fabric to be detected is abnormal according to the centroid distribution of each glue point in the target image; and obtaining a final fabric; and the application can greatly reduce the data processing amount and significantly improve the detection efficiency.

A method of encoding a video stream in a video encoder is provided that includes computing an offset into a transform matrix based on a transform block size, wherein a size of the transform matrix is larger than the transform block size, and wherein the transform matrix is one selected from a group consisting of a DCT transform matrix and an IDCT transform matrix, and transforming a residual block to generate a DCT coefficient block, wherein the offset is used to select elements of rows and columns of a DCT submatrix of the transform block size from the transform matrix.

The application discloses a low-quality Deepfake detection method based on high-frequency enhancement, and belongs to the technical field of Deepfake defense, and comprises the following steps: acquiring a face image set; extracting feature information, local high-frequency information and global high-frequency information of face images in the face image set; fusing the feature information, the local high-frequency information and the global high-frequency information respectively to obtain first fusion information and second fusion information; fusing the first fusion information and the second fusion information to obtain final fusion feature information; and inputting the final fusion feature information into a classifier for classification prediction to realize detection of a false face. According to the scheme, a local high-frequency information enhancement branch is constructed, a high-frequency amplitude enhancement is performed on DCT coefficients by using the characteristics of a channel attention mechanism, then a channel bottleneck module is constructed to remove coefficient noise, and a multi-space attention mechanism of frequency domain perception is indirectly realized through inverse DCT transformation.

The invention provides a manufacturing method, equipment and medium of a dispensing glue fabric, and relates to the technical field of dispensing glue fabric manufacturing, and the method comprises the steps: dividing a target image into a plurality of sub-images; performing DCT (Discrete Cosine Transform) on each sub-image to obtain a DCT coefficient matrix corresponding to each sub-image; according to the characteristics of each DCT coefficient matrix, all the DCT coefficient matrixes are clustered; processing each DCT coefficient matrix in the QA to obtain a processed DCT coefficient matrix cluster WA corresponding to the QA; mapping an energy center corresponding to each DCT coefficient matrix in the WA into the target image to obtain a mass center of a glue point corresponding to each DCT coefficient matrix in the QA; the mass center of a complete glue point composed of the incomplete glue points is obtained; according to the mass center distribution condition of each glue point in the target image, determining whether the to-be-detected point glue sol fabric is abnormal or not; the final fabric is obtained; according to the method, the data processing amount can be greatly reduced, and the detection efficiency is remarkably improved.

The application discloses a kind of toxic image generation method and black box classification model fingerprint watermarking method, it is related to neural network model protection and computer vision technical field, the method includes using DWT-DCT-SVD method to diffuse fingerprint watermark into whole image, and DCT coefficient is encrypted.To solve the problem of image quantity imbalance and class imbalance, a toxic feature enhancement module is introduced to improve the reliability and fidelity of the watermark. By combining the toxic feature enhancement module with the original classification module, a fingerprint watermark with low embedding strength can be effectively obtained. The fingerprint watermark in the toxic image generated by this method has excellent concealment. Even in the case of a 1-bit difference in the fingerprint watermark, copyright verification can still be accurately performed. Moreover, it shows strong robustness against various model watermark attacks.

Noise reduction in images is provided by performing a noise reduction step on blocks of pixels within a video-processing pipeline. The noise reduction step consists of applying a discrete cosine transform (DCT) to the block of pixels, quantizing the resulting DCT coefficients, and performing an inverse of the DCT to the quantized coefficients. The output of that noise reduction step is a block of image pixels similar to the input pixels, but with significantly less image noise. Because the noise reduction step can be performed quickly on small blocks of pixels, the noise reduction can be performed in real-time in a video processing pipeline.

The invention discloses a decoding method and electronic equipment. Belongs to the technical field of image processing. The embodiment of the method comprises the following steps: reading a first Huffman code of a first coding unit from a coding bit stream of a JPEG (Joint Photographic Experts Group) format image; a first additional bit length corresponding to the first Huffman code is inquired from a corresponding relation table, and the corresponding relation table is used for storing the corresponding relation between the Huffman code and the additional bit length; reading additional bit data from the coded bit stream based on the first additional bit length, and determining a discrete cosine transform (DCT) coefficient based on the first additional bit length and the additional bit data; and generating image data of the JPEG format image based on the DCT coefficient.

This invention discloses a high-capacity robust video watermarking method resistant to geometric attacks, comprising: acquiring the YUV three channels of a video; applying DTCWT transform to the U channel to separate the coefficients in each sub-band; selecting a watermark embedding strategy, dividing the coefficients into blocks and applying DCT transform, modifying the DCT coefficients according to the watermark information to be embedded, and sequentially applying inverse DCT transform and inverse DTCWT transform to obtain the watermarked U channel; combining it with the Y and V channels of the original video and performing the inverse YUV conversion operation, finally obtaining the watermarked video through video encoding. The extraction process is based on DCT transform, and the watermark is extracted according to the embedding strategy. This invention achieves robust watermark embedding and extraction based on DTCWT and DCT, flexibly balancing the robustness of the watermark against geometric attacks with the capacity of the payload, and can be applied to multimedia platforms where video is the mainstream.

The invention discloses a compressed domain video key frame extraction method and system based on a rough set theory. The method comprises the following steps of: extracting DCT (Discrete Cosine Transform) coefficients of all I frames of a video stream in a compressed domain, acquiring DC (Direct Current) coefficients, constructing an information decision table, calculating an average difference value of the DC coefficients between adjacent frames by utilizing a rough set theory, and eliminating redundant frames through attribute reduction to obtain a primary key frame set; and then optimizing the primary set, initializing a reference frame and a subsequent frame, adopting an SURF algorithm to extract local feature points and calculate similarity, judging content redundancy according to a preset threshold value, and further screening out a final key frame set. According to the method, the key frame is efficiently and accurately extracted in the compressed domain, the processing speed and the content representativeness are considered, and the problems that a traditional method is large in calculation overhead and high in redundancy are effectively solved.

The embodiment of the invention provides an image recompression method and related equipment. The method comprises the following steps of: obtaining a quantization discrete cosine transform (DCT) coefficient according to a first image; according to the context information of each quantized DCT coefficient, obtaining a probability distribution parameter corresponding to each quantized DCT coefficient; determining the probability of each quantized DCT coefficient according to the probability distribution parameter corresponding to each quantized DCT coefficient; and encoding the first image according to the probability of each quantized DCT coefficient. The proposed technology can save disk space.

The invention discloses an image watermark processing method, device and equipment, a medium and a product, and relates to the technical field of computer data security. The method comprises the following steps: acquiring a to-be-embedded image, watermark text information and a quantization step size; performing channel separation on the to-be-embedded image, and determining a first channel image, a second channel image and a third channel image; the first channel image is an R channel image; the second channel image is a G channel image; the third channel image is a B channel image; performing discrete cosine transform on the third channel image to obtain a first DCT coefficient matrix; performing binary conversion on the watermark text information to obtain watermark coding information; embedding the watermark coding information into an intermediate frequency region of the first DCT coefficient matrix to obtain a second DCT coefficient matrix; performing inverse discrete cosine transform on the second DCT coefficient matrix to obtain a fourth channel image; and obtaining a target image based on the fourth channel image, the first channel image and the second channel image. Therefore, the watermark processing efficiency is improved.

To meet the needs of hierarchical, multi-user, and differentiated applications in complex environments, this invention proposes a hierarchical image security and privacy protection method. This method divides the plaintext image to be encrypted into blocks, classifies the DCT coefficients of the image blocks, and uses compressed sensing technology for non-uniform compression. Combined with adaptive threshold sparsity and scrambling-diffusion encryption, it achieves progressive hierarchical image security and privacy protection for users with different access levels. Specifically, low-level users can only obtain the edge contour information of the image, medium-level users can obtain an image with blurred texture details, and high-level users can obtain an image containing both contour and detail information with the best visual effect, thus achieving hierarchical access control.