35 results about "Texture coding" patented technology

A system for encoding and compressing images and textures to a bitstream as an executable program of computer instructions operable on pixel space, permitting the Encoder to arbitrarily determine the order, location and method used for reconstructing the components of an image at the decoder. The bitstream is typically further compressed by an Adaptive Entropy Compressor. Bitstream execution reconstructs the image directly to pixel space. Memory management and parallel processing controls may be embedded in the bitstream. The system is optimised for AI processing of images with a universally interpretable construction format that enables machines and humans to understand precisely how the image was created and can be reconstructed. Dual bitstreams may be combined at the decoder for securely embedding customised content (e.g., advertising) at the edge.

The invention provides an industrial surface defect detection method fusing a large model and domain knowledge, and relates to the technical field of industrial detection, and the method specifically comprises the steps: designing a double-branch knowledge injection network model which comprises a backbone network, a side branch network, a texture encoder, a curvature extraction network and a detection head network, extracting general visual features by using a visual large model in the backbone network, then adding and fusing the general visual features with supplementary features extracted by the bypass network to obtain deep features, extracting texture features by using a texture encoder, and embedding texture domain knowledge into the deep features and the texture features in a cross attention mode to obtain a deep texture domain; optimizing a loss function according to the stress attention map; positioning and classifying defects by using a detection head network; and training and evaluating the double-branch knowledge injection network model. According to the technical scheme, the problems that in the prior art, the detection performance of a visual basic model in an industrial scene is reduced due to field differences, and the small sample generalization ability is weak are solved.

The present application relates to the fields of image processing and computer vision, and provides an image restoration method, device, electronic device, and computer storage medium. The method comprises: performing texture encoding on an input image through a texture encoding module, performing texture restoration on the encoded input image in combination with a texture branch network, and outputting a final texture restoration image; performing gradient encoding on the input image through a gradient encoding module, performing gradient restoration on the encoded input image in combination with a gradient branch network, and outputting a final gradient restoration image; performing super-resolution processing on the input image through a backbone network, and outputting a final reconstructed feature map; and fusing the final texture restoration image, the final gradient restoration image, and the final reconstructed feature map to obtain a restored image carrying texture information and gradient information. The image restoration method provided in the embodiment of the present application retains texture information and gradient information during image super-resolution restoration, thereby improving the restoration quality of low-resolution images.

The application discloses a kind of steam turbine generator unit shaft system three-dimensional real-time visualization monitoring method.It relates to the field of intelligent monitoring technology of electric power industry, method includes: using transfer matrix method, the vibration signal of sparse measuring point is as boundary condition, the displacement vector of whole shaft system discrete node and stress scalar are solved reversely, the global fidelity of physical data is established;Subsequently, through the construction of heterogeneous data channel of frequency domain decoupling and texture coding technology, physical data zero-copy is mapped to graphics rendering pipeline;Finally, in the vertex shading stage, introduce the double nonlinear model of sight distance sensitivity and space constraint, generate the dynamic gain factor strictly limited by physical gap, realize the conflict-free synthesis rendering of macroscopic rigid body displacement and microscopic elastic deformation in single window.The application realizes the physical real and visually readable shaft system digital twin monitoring by establishing the direct mathematical mapping from dynamics equation to graphics shader.

The utility model relates to a hard box cigarette anti -fake structure containing random texture distribution, aims at solving the problem of insufficient anti -fake effect, public identification difficulty and high cost in the prior art cigarette anti -fake technology. The anti -fake structure is arranged on the cigarette tongue of the cigarette hard box, contains the first material layer, second material layer and the random texture distribution layer between both. The first material layer is the transparent paper pulp layer or plastic film, ensures that the human eye can see the random texture pattern through it. The cigarette tongue or its package is equipped with the code, and the code corresponds with the random texture pattern and is stored in the database, and the consumer scans the code and compares and identifies. The cigarette tongue is also equipped with the layer knife line structure, and the first material layer is convenient for being opened to view the texture. The anti -fake structure adopts four layers and above paper pulp layer, and the texture distribution layer is located in the adjacent layer of the first material layer, prevents the imitation. The block breaking line is combined with the aluminum foil and is designed, and the block area is convenient for being extracted to identify. The utility model combines random texture, code database and multi -layer structure design, significantly enhances the anti -fake effect, reduces the cost, improves the public identification convenience, and has wide application prospect.

The invention provides a coding unit division method based on double texture complexity, and belongs to the technical field of video coding. Wherein the texture complexity of the target coding unit is detected, the unit type of the target coding unit is determined according to the texture complexity, a first preset threshold value and a second preset threshold value, and the unit type comprises an obvious texture coding unit or a non-obvious texture coding unit; according to the unit type of the target coding unit, a target division process of the target coding unit is determined, and the target division process comprises an obvious texture division process or a non-obvious texture coding division process; according to the target division process and the texture complexity, a target division strategy is determined, division operation is conducted on the target coding unit according to the target division strategy, and the target division strategy comprises a reverse division strategy, a quadtree combined multi-type tree division strategy or a standard division strategy. According to the invention, the effects of improving the video coding efficiency and improving the division efficiency and the division precision of the coding units can be achieved.

The invention relates to the technical field of intelligent buildings and elevators, and discloses a multi-dimensional track-based multi-compartment cooperative intelligent elevator system which comprises a multi-dimensional rack coding track system, at least one compartment unit and an intelligent scheduling system. The track system is formed by splicing rack track units, and racks used for driving meshing and texture codes containing absolute positions and track marks are integrated on the track system. The carriage unit is provided with a driving and positioning integrated assembly which comprises a driving gear meshed with the rack and an optical scanner for reading texture codes, and driving and positioning synchronization is achieved. And the intelligent scheduling system executes track prediction, conflict judgment and cooperative scheduling according to the accurate position reported by the compartment, and can instruct the compartment to perform transverse vacancy avoidance. According to the invention, through gear and rack forward meshing driving and track integrated coding positioning, high-precision and high-reliability cooperative operation of multiple carriages in a multi-dimensional track network is realized, and the traffic efficiency and scheduling flexibility in a building are improved.

Techniques are described for quickly finding a compression mode for BC7 encoding by modelling three sources of error in compression, namely, projection error, endpoint quantization error, and interpolation index quantization error. The mode with the lowest total error is selected for a block to be compressed.

Implementations relate to methods, systems, and computer-readable media for performing similarity analysis of three-dimensional (3D) objects. In some implementations, the method includes obtaining a 3D model of the object that includes a mesh and a texture, encoding the mesh into a mesh feature vector using a first neural network, encoding the texture into a texture feature vector using a second neural network, computing a mesh distance between the mesh feature vector and a reference mesh feature vector of a reference object, computing a texture distance between the texture feature vector and a reference texture feature vector of the reference object, determining, based on the mesh distance and the texture distance, whether the object matches the reference object, classifying the object as a dissimilar object if the object does not match the reference object, and classifying the object as a similar object if the object matches the reference object.

Techniques are described for training a machine learning (ML) model is learn compression errors for various compression modes of BC7 given an input set of features that depend on per-channel pixel value ranges in a BC7 block.

The application provides a multi-feature fusion-based no-reference screen image quality evaluation method and system, which comprises the following steps: calculating the gradient amplitude, relative gradient amplitude and gradient direction mapping of the input screen content image, applying a local ternary pattern operator on the mapping for texture coding, combining the gradient domain values to statistically weight the coding modes, and generating a gradient-weighted local ternary pattern histogram; inputting the screen content image into a pre-trained deep convolutional neural network for feature extraction, obtaining the deep feature map at the end of the network, and converting the deep feature map into a global perception feature vector through a global average pooling operation; fusing the gradient-weighted local ternary pattern histogram and the global perception feature vector, inputting them into a neural network quality prediction model, and outputting the no-reference quality evaluation score of the screen content image after nonlinear mapping of the model. The application realizes accurate mapping from multi-dimensional features to quality scores.

A picture streaming method performed by an electronic device includes obtaining texture data of a virtual scene picture of a target application based on a target streaming manner selected from a plurality of streaming manners. The target application is configured to render the virtual scene picture. The method further includes encoding the texture data to obtain a texture encoding result, and transmitting the texture encoding result to a head mounted display device. The texture encoding result is configured for triggering the head mounted display device to decode the texture encoding result to obtain the texture data, and display the virtual scene picture based on the texture data.

The present application belongs to the technical field of image processing, and relates to a compressor cast iron surface defect recognition method based on multi-scale texture analysis. The method slides a detection window on the gray image of the surface of the cast iron to be detected, constructs a gradient histogram and calculates a knife mark concentration to determine the window class; a multi-scale binary texture coding is performed with the main gradient direction as the starting angle, the normalized texture entropy value of each scale is calculated as the entropy anomaly, and the main inspection scale and the main entropy anomaly are selected based on the window class; the proportion of the normalized uniform mode under each scale is calculated to obtain the uniform anomaly and the main uniform anomaly; finally, in response to the main entropy anomaly and the main uniform anomaly satisfying the dynamic first preset threshold and the second preset threshold respectively, the defect recognition result is output. The present application effectively avoids the dilution of the background features to the defect signal, and improves the sensitivity of the micro defect detection and the system noise resistance under the complex background of multi-state interweaving.

The invention discloses a pedestrian detection-oriented physical adversarial patch generation and migration method, and belongs to the field of computer vision and artificial intelligence security. The invention provides a shape-texture two-stage layered optimization and Min-Max robust migration framework aiming at the problems of abrupt patch shape, difficulty in reproduction of pixel-level noise and poor environmental adaptability in the existing physical countermeasure attack. The method comprises the following steps: firstly, constructing a smooth patch contour by using a centripetal Catmull-Rom spline curve, and optimizing an anchor point coordinate through a differential evolution algorithm so as to lock an effective receptive field which is most sensitive to a model; secondly, constructing a discrete texture coding space based on a physical adhesive tape, and searching an optimal stripe texture formed by the color, width and direction of a standard adhesive tape in a fixed shape; meanwhile, a Min-Max robust optimization strategy is introduced, and the anti-interference capability of the patch is improved by minimizing the maximum confidence coefficient under the worst illumination transformation; and finally, a physical mapping strategy based on vector reconstruction is adopted, a precise transformation model from pixels to millimeters is established, and an SVG vector manufacturing file irrelevant to resolution is output. According to the method, precise closed loop of the anti-patch from digital algorithm optimization to physical object manufacturing is realized, and the method has the advantages of high physical realizability, natural and hidden shape, high environmental robustness and the like.

The invention relates to the technical field of image processing, and discloses a cross-graph context-aware cell nucleus instance segmentation method and system, and the method comprises the steps: obtaining a cell image, segmenting the cell image into slice groups, and constructing a cell nucleus instance segmentation model which comprises a feature extraction module, a context injection module, a prompt coding module, a mask decoder and a texture encoder; a feature extraction module extracts prior features and domain features of the slices, refines the prior features through a multi-stage feature refinement block, and extracts and fuses missing features from the domain features to obtain enhanced features; a context injection module integrates context information according to the enhanced features to generate image features, a prompt coding module generates point embedding, a mask decoder obtains a mask of each cell nucleus according to the image features and the point embedding, and the masks are combined to obtain an instance segmentation result; a texture encoder encodes the enhancement feature and the instance segmentation result into a texture feature, which is stored in a context injection module together with the enhancement feature. The embodiment segmentation effect can be improved.

The application relates to the field of intelligent decision-making, and discloses a face living body detection method, device and equipment in a motion state scene and a medium. The method comprises the following steps: collecting a face image in a motion state scene, performing face detection on the face image to obtain a detected face, performing format standardization on the detected face to obtain a standardized face; performing texture coding on the standardized face to obtain coded texture of the standardized face, extracting texture information, calculating pixel change information, and constructing three-dimensional structure information of the standardized face; respectively performing feature extraction on the texture information, the pixel change information and the three-dimensional structure information to obtain texture features, pixel change features and three-dimensional structure features; performing feature fusion on the texture features, the pixel change features and the three-dimensional structure features to obtain fused features; and calculating a living body detection score of the standardized face to determine a face living body detection result of the face image. The application can improve the comprehensiveness of face living body detection in a motion state scene.

This invention belongs to the field of image processing technology and relates to a method for identifying surface defects in compressor cast iron parts based on multi-scale texture analysis. The method involves sliding a detection window on a grayscale image of the cast iron part surface to be inspected, constructing a gradient histogram, and calculating the concentration of tool marks to determine the window category. Multi-scale binary texture encoding is performed using the principal gradient direction as the starting angle. The normalized texture entropy value at each scale is calculated as the entropy anomaly, and the principal inspection scale and principal entropy anomaly are selected based on the window category. The proportion of uniform patterns after encoding normalization at each scale is calculated to obtain the uniform anomaly and the principal uniform anomaly. Finally, in response to the principal entropy anomaly and the principal uniform anomaly satisfying a dynamic first preset threshold and a second preset threshold, respectively, the defect identification result is output. This invention effectively avoids the dilution of defect signals by background features and improves the sensitivity and noise resistance of detecting small defects in complex, multi-modal backgrounds.

The application relates to a design method of a city population flow visualization system, which comprises the following steps: inputting city coordinate sample point data into a newly created point texture, and inputting the city coordinate sample point data into a fragment shader of a GPU to obtain a velocity field of the city coordinate sample point; constructing a Voronoi polygon grid in which or around which a preset feature point exists in a unit grid in the GPU with the city coordinate sample point as the center; calculating the distance and comprehensive influence intensity of any pixel point in the fragment shader to all preset feature points to generate a density field with different intensities; performing RGBA texture encoding on particle coordinates in a particle rendering system, decoding and calculating the current position of the particle in the fragment shader; generating different numbers of particles according to the density field with different intensities, and showing the net inflow of the city population by the number of the particles; and updating the particle position according to the velocity field to realize a population flow effect. The application realizes high-performance visualization display of population flow between cities.

This application provides a coding unit partitioning method based on dual texture complexity, belonging to the field of video coding technology. The method involves: detecting the texture complexity of the target coding unit; determining the unit type of the target coding unit based on the texture complexity, a first preset threshold, and a second preset threshold; the unit type including: a coding unit with obvious texture or a coding unit without obvious texture; determining the target partitioning process for the target coding unit based on the unit type; the target partitioning process including: a partitioning process with obvious texture or a partitioning process without obvious texture; determining the target partitioning strategy based on the target partitioning process and texture complexity; and performing the partitioning operation on the target coding unit according to the target partitioning strategy, which includes: a reverse partitioning strategy, a quadtree combined with multiple tree types partitioning strategy, or a standard partitioning strategy. This application can improve video coding efficiency and the partitioning efficiency and accuracy of coding units.

The present application relates to the technical field of image processing, and discloses a cross-image context-aware nucleus instance segmentation method and system, comprising: acquiring a cell image and segmenting it into a slice group, and constructing a nucleus instance segmentation model comprising a feature extraction module, a context injection module, a prompt encoding module, a mask decoder and a texture encoder; the feature extraction module extracts prior features and domain features of the slice, refines the prior features through a multi-level feature refinement block, extracts and fuses missing features from the domain features to obtain enhanced features; the context injection module generates image features by integrating context information according to the enhanced features, the prompt encoding module generates point embedding, the mask decoder obtains a mask of each nucleus according to the image features and the point embedding, and the instance segmentation result is obtained by combining the masks; the texture encoder encodes the enhanced features and the instance segmentation result into texture features, which are stored into the context injection module together with the enhanced features. The present application can improve the instance segmentation effect.

Techniques are described for quickly finding a compression mode for BC7 encoding by modelling three sources of error in compression, namely, projection error, endpoint quantization error, and interpolation index quantization error. The mode with the lowest total error is selected for a block to be compressed.

The invention provides an image loading method and electronic equipment. The electronic equipment can perform image decoding on an original image and scale the original image to generate a bitmap with a preset size. The electronic equipment can perform texture coding on the bitmap with the preset size to obtain a texture thumbnail with the preset size. The electronic equipment can store the texture thumbnail and load the texture thumbnail for image display when the grid page is displayed. The GPU of the electronic device may identify and use the texture thumbnail. After the electronic equipment loads the texture thumbnail, image decoding does not need to be carried out. Therefore, the time required for displaying the image on the grid page can be shortened, and the smoothness of browsing the image on the grid page in a sliding manner is improved.

The application provides a generative sparse latent color field method for three-dimensional original texture generation, comprising: acquiring geometric information of a target three-dimensional grid to form point cloud data; inputting the point cloud data into a pre-trained texture variational autoencoder to generate geometric conditioned latent features aligned with a texture latent space; based on a geometric conditioning generation model, sampling and generating a latent feature representation of a target texture in a sparse latent space with the geometric conditioned latent features as conditions; inputting the latent feature representation of the target texture into a pre-trained continuous color field decoder to query and decode surface points of the target three-dimensional grid to obtain seamless texture applied to the target three-dimensional grid. The problem of inconsistent geometry / texture coding in the existing projection method is solved.

The invention provides an image restoration method. The method comprises the following steps: acquiring a to-be-restored image, a corresponding mask image, a sheltered edge image and a sheltered grayscale image; obtaining pre-trained image restoration (DESSM) based on a state space model; inputting the sheltered image and the mask image into a texture encoder in the DESSM after passing through an embedding layer, and inputting the sheltered edge image, the sheltered gray level image and the mask image into a structure encoder in the DESSM after passing through the embedding layer; the decoder outputs the restored image. According to the method, the DESSMs based on the state space model are creatively provided, the global dependence of the defect image is effectively captured, meanwhile, the linear complexity is kept, the method is superior to the current most advanced image restoration models based on CNN, Transformers, a pre-training diffusion model and the like, and the overall performance is proved to be obviously improved.

The invention belongs to the field of deep learning and data coding, and provides a depth feature map and image texture-oriented collaborative coding method, which is used for collaboratively reconstructing a high-quality image while efficient feature coding is carried out, and meeting machine vision tasks and human vision requirements at the same time. According to the method, a feature coding and texture coding double-branch collaborative framework is constructed, during feature coding, a feature map which is not highly related to a specific visual task is removed through a feature map selection module, and the bit number required by feature map data transmission is remarkably reduced while the downstream task performance is ensured; meanwhile, a low-resolution image decoded by a texture branch is used for assisting in restoring a missing part in a feature map, a high-quality feature map obtained after feature map branch enhancement is used as prior information, image super-resolution reconstruction of the texture branch is guided, and by means of the two-way cooperation mechanism, super-resolution reconstruction of the texture branch is achieved. And the representation capability of the final feature map and the visual quality of the reconstructed image are effectively enhanced.

The application provides an industrial surface defect detection method fusing a large model and field knowledge, relates to the technical field of industrial detection, and specifically comprises the following steps: designing a double-branch knowledge injection network model, including a backbone network, a side branch network, a texture encoder, a curvature extraction network and a detection head network; extracting general visual features by using a visual large model in the backbone network, then adding and fusing the general visual features and supplementary features extracted by the side branch network to obtain deep features; extracting texture features by using the texture encoder; embedding texture field knowledge by using the deep features and the texture features in a cross attention manner; and optimizing a loss function according to a stress attention graph; positioning and classifying defects by using the detection head network; and training and evaluating the double-branch knowledge injection network model. The technical scheme of the application overcomes the problems of the existing technology, such as the decline of detection performance in an industrial scene caused by field differences and the weak small sample generalization ability of a visual basic model.

The invention provides a picture display method and a related device, in the method, when an electronic device stores a thumbnail corresponding to a shot or saved image, the electronic device can firstly perform texture coding on the thumbnail to obtain texture data. Then, the electronic equipment carries out hypercompression on the obtained texture data to obtain hypercompressed texture data; when the electronic device loads and displays a thumbnail based on a user operation, the electronic device firstly obtains the hypercompressed texture data. And the electronic equipment decodes the supercompressed texture data into texture data, and splices the texture data corresponding to the plurality of thumbnails into a large texture graph. And the electronic equipment renders and displays the large texture graph. In this way, the texture data is stored locally after being supercompressed, so that the space occupied when the thumbnail is stored locally can be reduced, and the consumption of the application to the memory of the whole machine is reduced. In the process of loading the thumbnail by the electronic equipment, the supercompressed data is read, so that the IO overhead during batch data reading is reduced.

In various embodiments, a training application trains a machine learning model to generate three-dimensional (3D) representations of two-dimensional images. The training application maps a depth image and a viewpoint to signed distance function (SDF) values associated with 3D query points. The training application maps a red, blue, and green (RGB) image to radiance values associated with the 3DI query points. The training application computes a red, blue, green, and depth (RGBD) reconstruction loss based on at least the SDF values and the radiance values. The training application modifies at least one of a pre-trained geometry encoder, a pre-trained geometry decoder, an untrained texture encoder, or an untrained texture decoder based on the RGBD reconstruction loss to generate a trained machine learning model that generates 3D representations of RGBD images.

Techniques are described for training (2208) a machine learning (ML) model that learns compression errors for various compression modes of BC7 given an input feature set that depends on a range of pixel values per channel in the BC7 block.