28 results about "Texture synthesis" patented technology

The invention discloses an ancient textile image restoration system based on artificial intelligence. The system comprises a multi-source image acquisition module, a damaged area detection module, a pattern generation module, a color restoration module, a texture synthesis module and a multi-scale fusion module. The system introduces a wavelet guidance-frequency domain attention mechanism and a rotation invariant Haar wavelet basis function to realize accurate identification and classification of a damaged area; a saliency-guided wavelet decomposition control and self-adaptive threshold denoising method is combined, so that the perception capability of slant textures and edge details is improved; the texture synthesis module constructs a hierarchical modeling strategy fusing Gram style loss, Wasserstein style loss and a total variation regular term, and realizes generation of high-quality textures with unified styles and smooth edges; the system can be widely applied to cultural relic digital repair and display scenes.

The invention relates to a raised dust pollution source image enhancement processing method and system, and the method specifically comprises the following steps: collecting a raised dust pollution source image and environment parameters, and marking a raised dust region and a background to construct a data set; estimating image illumination and reflection components by using multi-scale Gaussian kernel convolution, fusing the image illumination and reflection components to obtain an enhanced image, enhancing the contour through adaptive morphological operation, and generating a morphological enhanced image; real raised dust texture is injected through random deformation field elastic transformation in combination with texture synthesis, and a synthetic enhanced image is obtained; constructing an image enhancement model, inputting a synthetic enhanced image, extracting multi-scale features by an encoder, enhancing the features by a decoder, reserving boundaries, optimizing the model in combination with a loss function, enhancing an output result through region self-adaptive fusion, and completing model training; and finally, inputting a newly acquired image into the trained model to generate a final enhanced flying dust pollution source image. The flying dust image quality is improved, features are accurately extracted, and pollution source identification and monitoring are assisted.

The invention belongs to the technical field of material image recognition and processing, and provides a metallographic microscopic image structure classification method based on multi-scale structure feature extraction. The method comprises the steps of microscopic image acquisition, preprocessing, defect suppression, multi-scale structural feature extraction, feature fusion, dimension reduction and classification judgment, integration of grain boundary, texture and morphological features, defect repair by adopting PatchMatch texture synthesis, and reduction of the influence of scratches, uneven corrosion and the like on precision. Feature fusion supports weight adaptive distribution, and a classifier can be a support vector machine (SVM) or a lightweight convolutional neural network (preferably MobileNet). The system is composed of an acquisition module, a preprocessing module, a defect detection and repair module, a feature extraction module, a fusion and dimension reduction module and a classification judgment module, can be in butt joint with a required platform, realizes online detection, automatic classification and result archiving, is suitable for metal and non-metal material microscopic structure classification, and has a good application prospect. Wherein the non-metallic material can increase porosity calculation, fiber direction distribution and other morphological characteristic analysis.

The invention relates to the field of remote sensing image processing, in particular to a remote sensing image super-resolution reconstruction and detail enhancement deep learning system which comprises a super-resolution reconstruction network module, a texture synthesis detail enhancement network module based on manifold learning, a training management module and a preprocessing module. Wherein the super-resolution reconstruction network module adopts an architecture based on a generative adversarial network, and provides progressive 1 / 4, 1 / 2 and 1-time resolution reconstruction; the manifold learning-based texture synthesis detail enhancement network module extracts texture features through multi-level high-pass filtering, maps the features to a manifold space, generates a texture residual error by adopting a geodesic interpolation principle, and applies manifold constraint optimization for a specific ground feature category so as to obtain a manifold learning-based texture synthesis detail enhancement network. According to the system, two tasks of super-resolution reconstruction and detail enhancement are innovatively combined, more accurate texture feature capture and reconstruction are realized, and the adaptability of the system to different surface feature categories is improved.

The invention discloses a two-dimensional code recognition and defect repair method and device, and the method achieves the effective integration of visible light, infrared and ultraviolet image features through the collection and preprocessing of a multi-modal image, greatly improves the feature recognition degree of a two-dimensional code under a complex background, and solves a problem that a conventional method is poor in adaptability to the complex background. According to the lightweight convolutional neural network defect detection model, pixel-level defect segmentation and multi-dimensional classification are realized, a defect severity grading mechanism is combined, an accurate basis is provided for intelligent defect evaluation, and the defect of lack of intelligent evaluation in the prior art is made up; for self-adaptive repair strategies of defects of different types and levels, the advantages of multiple algorithms such as texture synthesis, error correction coding and geometric correction are fused, and the repair precision of the two-dimensional code is greatly improved; a closed-loop verification mechanism of repair and recognition realizes linkage optimization of a repair process and a recognition result by dynamically adjusting repair parameters, and the problem of low efficiency caused by separation of repair and recognition is effectively solved.

A method includes generating a three-dimensional (3D) mesh of the virtual environment from a set of images of physical terrain and depth information. The method further includes applying an initial texture to the 3D mesh based on the depth information. The method further includes generating a projection of a 360-degree panorama onto the 3D mesh. The method further includes identifying one or more areas in the projection that are missing the initial texture. The method further includes providing the projection, identification of the one or more areas in the projection that are missing the initial texture, and the depth information to a machine-learning model. The method further includes outputting output texture for the one or more areas in the projection that are missing the initial texture and blending the output texture with the initial texture in the projection to obtain a blended texture in the projection.

The invention relates to a facade texture repairing method and device, electronic equipment and a computer storage medium, and belongs to the technical field of geographic information surveying and mapping, and the facade texture repairing method comprises the steps: obtaining a multi-view texture image of a building facade, and determining a main texture image and an auxiliary texture image in the multi-view texture image; the view angle of the main texture image is taken as a main view angle, the auxiliary texture image is converted to the main view angle by adopting a homography matrix, a candidate texture image is obtained, and covered areas of the main texture image and the candidate texture image are determined based on the structural similarity of the main texture image and the candidate texture image; and calculating a cost value during texture synthesis of the covered area based on the main texture image and the candidate texture image by adopting a preset texture synthesis cost function, determining a texture synthesis mode with the minimum cost value, and obtaining a texture restoration image of the building facade. According to the invention, accurate texture repair of the building facade can be realized.

Systems and methods are provided herein for efficient, reproducible noise sampling coordinate generation for texture synthesis in image and video processing. The techniques enable high-quality film grain or texture synthesis as a post-processing step in video decoders, masking compression artifacts and restoring perceptual quality. Hash-based mapping, including direct coordinate mapping and linear index hashing, can be used to generate coordinates for noise sampling by hashing image block coordinates and a seed with preselected constants to generate sampling positions. Shuffled permutation sampling can be used to generate coordinates by generating and shuffling a list of template coordinates using a seeded pseudo-random number generator, and assigning unique positions to each block. Wrap-around addressing, including treating the noise template as a toroidal space, can be used to maximize coverage and eliminate edge bias. The disclosed techniques provide deterministic, hardware-friendly, and unbiased sampling.

The present application relates to a kind of dust pollution source image enhancement processing method and system, specifically as follows: collecting dust pollution source image and environmental parameter, dust area and background are labeled to construct data set;With multiscale Gaussian kernel convolution estimation image illumination, reflection component and fusion are enhanced, then contour is strengthened by adaptive morphological operation, and morphological enhancement image is generated;Real dust texture is injected by random morphing field elastic transformation combined with texture synthesis, and synthesis enhancement image is obtained;Image enhancement model is constructed, input synthesis enhancement image, multiscale feature is extracted by encoder, decoder enhances feature and retains boundary, loss function is combined to optimize model, and the output result is also enhanced by region adaptive fusion, complete model training;Finally, new acquisition image is input into trained model, and the final enhanced dust pollution source image is generated.The present application improves dust image quality, accurately extracts feature, and helps pollution source identification and monitoring.

The invention discloses a wall painting digital restoration method based on layered line draft constraint, which comprises the following steps of: decomposing a wall painting line draft into a contour layer, a detail layer and a texture layer, extracting visual information of each layer through multi-scale analysis, and performing layered restoration by respectively adopting a structure propagation algorithm, a texture synthesis algorithm, a color migration technology and the like. In the restoration process, each layer of line draft serves as a constraint condition to guide the structure, detail and texture restoration of a restoration area, a result is refined step by step in combination with an iterative optimization strategy, global optimization based on the deep learning technology is carried out, and finally a high-precision restoration image is output. The innovation point of the method lies in a layered line draft constraint mechanism, a targeted repair algorithm based on the line draft and deep learning technology auxiliary optimization, the repair precision, efficiency and naturalness of the complex damaged mural are remarkably improved, and the method is suitable for the fields of cultural relic protection and cultural heritage digitization.

Various examples, systems, and methods are disclosed relating to texture synthesis. A first computing system determine, using a denoiser and based at least on an input indicating one or more characteristics of a scene, a plurality of estimated views of the scene corresponding to a texture. The first computing system can render, from a model of the texture, a plurality of renders of the texture, at least one render of the plurality of renders being associated with a corresponding estimated view of the plurality of estimated views. The first computing system can update the model of the texture based at least on the plurality of renders and the plurality of estimated views. The first computing system can update the plurality of estimated views based at least on the plurality of renders.

This invention discloses a seamless removal and restoration method and system for specific targets in remote sensing images. Addressing the problems of high redundancy in horizontal detection box masks and poor structure preservation in restoration models in existing technologies, this invention first utilizes an improved YOLOv11_obb network to perform rotating target detection on the remote sensing image, obtaining rotated bounding boxes with angle information. Secondly, a binarized rotated mask is generated based on the rotated bounding boxes to preserve the original background information to the greatest extent. Finally, the remote sensing image and the mask are stitched together and input into a network restoration model based on the LaMa architecture. Leveraging the global receptive field characteristics of FFC, long-range dependencies are captured in the frequency domain, and texture synthesis and structural reconstruction are performed on the masked region to obtain a seamless restored image after the removal of the specific target. This invention can effectively remove specific targets such as aircraft and ships, and the restored image significantly outperforms existing diffusion models and traditional restoration methods in terms of geometric structure consistency and color fidelity, making it suitable for scenarios such as privacy protection and feature editing.

The invention discloses a cartoon character feature mapping method based on image recognition and affine transformation, and relates to the technical field of computer vision, image processing and three-dimensional modeling. The method specifically comprises the steps of image input, key feature detection and segmentation, eye center coordinate calculation, UV expanded image coordinate alignment and affine transformation, chartlet generation and texture synthesis, and three-dimensional rendering output. The invention discloses an intelligent mapping method and system for cartoon character features on a 3D model car window based on image recognition and affine transformation, and solves the problems that cartoon image features are difficult to automatically recognize, the mapping precision is low, and three-dimensional rendering is not matched in the prior art. The intelligent and high-precision feature mapping and rendering from the cartoon image input to the three-dimensional model vehicle window area are realized.

A method includes generating a three-dimensional (3D) grid of a virtual environment from an image set of physical topography and depth information. The method further includes applying the initial texture to the 3D mesh based on the depth information. The method further includes generating a projection of the 360-degree panorama onto the 3D mesh. The method further includes identifying one or more regions in the projection lacking the initial texture. The method further includes providing the projection, identification of the one or more regions in the projection lacking the initial texture, and depth information to the machine learning model. The method further includes outputting an output texture of the one or more regions in the projection lacking the initial texture, and mixing the output texture with the initial texture in the projection to obtain a mixed texture in the projection.

The invention provides a mesoscale appearance synthesis rendering method and system based on three-dimensional Gaussian splashing, and relates to the technical field of computer graphics and real-time rendering. Comprising the following steps: acquiring a multi-view picture containing a mesoscale appearance object, and extracting a basic grid; the three-dimensional gauss are stored as textures in a layered mode according to the height; mapping the multi-layer three-dimensional Gaussian texture to the basic grid to obtain a rendering result; calculating loss according to the rendering result and the corresponding view angle picture, and training a multilayer three-dimensional Gaussian texture; based on a two-dimensional image synthesis method, performing Gaussian texture synthesis on the trained multi-layer three-dimensional Gaussian texture to obtain a synthesized Gaussian texture; and mapping the synthesized Gaussian texture to a corresponding new object, and rendering to obtain a mesoscale appearance under new illumination. According to the method, the mesoscale appearance is represented by using the multi-layer three-dimensional Gaussian texture modeling, more mesoscale structure details are expressed, the precision of the mesoscale appearance constructed on the new object is improved, and real-time synthesis rendering of the Gaussian texture on the new object is realized.

The invention relates to a remote sensing scene new visual angle synthesis method based on depth supervision and a texture discriminator. The method depends on implicit scene representation of a neural radiation field and a differentiable rendering technology, and is a new visual angle synthesis method of a remote sensing scene under a sparse input condition. In the neural radiation field optimization process, a depth supervision optimization framework based on sparse point cloud is introduced, and model overfitting is avoided by aligning depth information at key points and correcting scene geometry. In addition, a texture discriminator trained based on a generative adversarial mode is introduced, and the texture synthesis quality of the occlusion area is effectively improved by constructing the perception of a positive and negative sample pair improvement model on a scene. According to the method, a high-quality new view angle image synthesis effect can be realized under sparse view angle input, the defects of a current deep learning method in the aspect of new view angle synthesis research of a remote sensing scene are overcome, and the method has wide application prospects and application values.

An image processing system includes a material image generator, at least one texture generator, and an output controller. The material image generator is configured to generate a material image. The at least one texture generator is coupled to the material generator unit and configured to adjust texture characteristics of the material image to generate at least one texture image. The output controller is coupled to the at least one texture generator and configured to perform region characteristic analysis on a source image to generate analysis results, determine region weights based on the analysis results, and synthesize the at least one texture image with the source image based on the region weights to generate an output image.

The invention provides a multi-view texture synthesis method and device, electronic equipment and a storage medium. A vertex view weight is constructed by synthesizing an Euclidean distance and a sight line included angle on a vertex level; the vertex weight is interpolated into a sampling point level view mixed weight by using the barycentric coordinates of texture sampling points relative to a target triangular patch in a texture parameter domain, and weighted fusion is realized by combining visibility suppression, weight normalization and multi-view projection interpolation sampling; therefore, the fusion weight is continuously changed in the surface patch, the view jump and the splicing artifact of adjacent sampling points are reduced, the continuity and the consistency of texture synthesis are improved, meanwhile, the ineffective view interference is inhibited under the condition of shielding or unfavorable view angle, a more stable and more natural final texture image is obtained, and packaging and exporting of texture blocks are facilitated.

This invention discloses a texture synthesis method based on Maya 3D model data, belonging to the field of texture synthesis technology. The method includes the following steps: model creation, 3D model parameterization, texture matching, using a texture detection algorithm in the 3D model data texture synthesis system to extract texture structure features from a texture sample image, obtaining a texture structure feature image using the texture detection algorithm, texture synthesis, selecting a rectangular texture block from the texture sample image as the first texture block of the texture image, globally scanning the texture sample image to minimize the error in the overlapping portion between the selected texture block and the first texture block, comparing the structural feature images of the texture blocks to ensure the overlapping area structures match, and then cutting and stitching to obtain a complete texture image. This invention avoids the situation where the texture structure is completely mismatched due to pixel error values ​​in the overlapping area during texture block matching, and can synthesize more detailed texture images.

The present application relates to the fields of finance, medical health and artificial intelligence technology, and provides a face fusion method, device, equipment and medium based on parallel denoising, which can detect face key point coordinates in a target image based on a face key point detector, extract identity features of a source face image based on an identity encoder, calculate texture features of the source face image based on a texture statistics extractor, so as to provide a data basis for subsequent fusion; inject the identity features into a denoising process based on the face key point coordinates using an identity maintaining path, generate skin texture and illumination details using a texture synthesis path, decompose a single path diffusion model into two parallel and function-specific processing paths, so as to avoid coupling conflicts when simultaneously optimizing identity and texture; and fuse output features of the identity maintaining path and the texture synthesis path using a dynamic gating module, so that the fusion strategy can adapt to the needs of different denoising stages, thereby generating a high-quality face fusion image.

The invention relates to the technical field of metallographic analysis and grain size detection of metal materials, in particular to a twin boundary detection and elimination method for intelligent grading of grain sizes, and aims to solve the problem of deviation of an automatic grading result of the grain sizes due to the fact that twin boundaries generated after materials such as austenitic stainless steel or copper alloy are etched are easily misjudged as normal grain boundaries. According to the method, precise detection of the twin boundaries is realized through Hough transform, image restoration after twin boundaries are removed is completed in combination with a texture synthesis technology based on a Criminisi algorithm, interference of the twin boundaries on intelligent grain size rating is effectively solved, the processed metallographic image can be directly used for subsequent automatic grain size calculation, and rating precision is remarkably improved.

The present disclosure provides a multi-view texture synthesis method and device, electronic equipment and storage medium. The vertex view weight is constructed by comprehensively constructing the Euclidean distance and the view angle at the vertex level, and the vertex weight is interpolated into the view mixing weight at the sampling point level by using the barycentric coordinates of the texture sampling point relative to the target triangular patch in the texture parameter domain. Then, the weighted fusion is realized by combining the visibility suppression, weight normalization and multi-view projection interpolation sampling, so that the fusion weight continuously changes in the patch, reduces the view jump and seam artifacts of adjacent sampling points, improves the continuity and consistency of texture synthesis, and suppresses the invalid view interference under the conditions of occlusion or unfavorable view angle. Thus, a more stable and natural final texture map is obtained, and the texture block is packaged and exported.

A computer-implemented method for generating a target texture (201) from at least two source textures (211, 221, 231) includes performing a statistical texture synthesis operation. An interpolation weight is assigned to each of the source textures, and the target texture is synthesized using the source textures and the assigned interpolation weights. Synthesizing the target texture includes: randomly selecting one of the source textures; randomly extracting a texture patch (212) from the selected source texture (211); modifying the extracted texture patch in such a way that at least one statistical property (215) of the modified texture patch (214) approximates a corresponding average statistical property (202); and inserting the modified texture patch into the target texture such that the modified texture patch seamlessly fits the existing texture content in the target texture (203).

The invention discloses a cross-application content display decoupling method and system based on Activity View, and relates to the technical field of computer software and mobile terminals.The method comprises the steps that an Activity View container is instantiated by a host application program, and virtual display equipment associated with a graph buffer area is created through a management subsystem; requesting to run an export component of the client application on the virtual display device; the window manager draws a client interface to a graph buffer area; the graphic synthesizer reads data in the buffer area and remaps and synthesizes the data to a container area of a physical screen, the host intercepts an input event, coordinate transformation is carried out based on the geometric mapping relation, and a target event is generated and injected into a client input channel. Through the virtual display and texture synthesis technology, process isolation and UI seamless integration of the host and the client are achieved, cross-process interaction and life cycle synchronization are supported, and the decoupling performance and safety of the system are improved.

The invention discloses a texture structure design space, a design method and a design system of a flexible structure for 3D printing and the flexible structure. The design method comprises the steps of designing spiral texture structure units, constructing the texture structure design space, splicing the corresponding spiral texture structure units in a grid surface, adjusting local characteristics of the flexible structure and forming the final flexible structure. A spiral-based elastic structure unit is used as a basic construction block for texture synthesis to generate a self-adaptive texture structure, so that the whole working process from design to manufacturing is simplified, an interactive user interface is adopted, texture design, performance preview and G code generation are seamlessly integrated together, cross-platform conversion is not needed, cross-platform conversion is eliminated, and the working efficiency is improved. And the manufacturing efficiency is obviously improved. A data-driven prediction model is adopted, and key mechanical characteristics such as rigidity and ductility can be efficiently approximated. The prediction result is directly visualized on the design structure, visual feedback is provided, and seamless iterative design optimization can be achieved.

The present application belongs to the field of natural image processing, and particularly relates to a structure and texture mutual promotion image repairing method. The present application follows the basic framework of a generative adversarial network, and comprises two parts of a double-branch parallel coupling generator and a double-branch discriminator. The double-branch parallel coupling generator utilizes two branches to respectively reconstruct a structure and synthesize a texture, in the process, the two branches frequently interact in both directions, so that the structure and the texture acquire complementary information from each other, form a mode of structure guiding texture synthesis and texture promoting structure reconstruction, and meanwhile, the frequent interaction of the two branches realizes mutual perception of the structure and the texture, and accurately ensures the consistency between the two. The double-branch discriminator discriminates the generated image of the generator from the structure and the texture, contains two processes of structure and texture adversarial, and supervises the generator to generate an image with reasonable structure and texture.