420 results about "Inpainting" patented technology

The invention provides an image restoration and super-resolution reconstruction system and method based on deep learning, and belongs to the technical field of digital image processing. The invention aims to solve the problems of high calculation complexity and resource consumption, limitation of long sequence processing, high training difficulty and texture scene deficiency when a multi-scale residual network based on a Transform architecture is used for image resolution conversion. The reconstruction system comprises: an image preprocessing module performing window division and video memory optimization on an input low-resolution image; the multi-layer fusion network dynamically adjusts the characteristics of the low-resolution image, captures channel information in different scenes, performs interactive fusion, performs comparison supervision, establishes an information communication channel, dynamically adjusts and optimizes parameters through negative feedback, and obtains a super-resolution image. And the loss function module maximizes the similarity of the super-resolution image and the high-resolution image in the segmentation feature space to obtain a final super-resolution image.

The invention discloses a three-dimensional scene generation method based on a diffusion model and Gaussian splashing and related equipment, and belongs to the field of computer vision. The method comprises a point cloud construction stage and a three-dimensional Gaussian optimization stage: in the point cloud construction stage, extracting an initial point cloud from an input initial image, and expanding the point cloud through a Stable Diffusion image restoration model, a monocular depth estimation model and a boundary perception depth alignment module to create a scene; in the three-dimensional Gaussian optimization stage, three-dimensional Gaussian is initialized according to the obtained point cloud, the three-dimensional Gaussian is utilized to represent the whole scene, a multi-view image is adopted as a supervision signal to optimize parameters of the three-dimensional Gaussian, and a three-dimensional scene capable of being browsed in an immersive mode is obtained after optimization. According to the method, the high-fidelity three-dimensional scene with accurate geometry and natural appearance can be generated without training any three-dimensional data, the problem of dependence on three-dimensional scene data is solved, the requirement for computing power resources is small, and the manufacturing cost of the three-dimensional scene is reduced.

The invention provides a blind person face image restoration method and apparatus, an electronic device and a storage medium. The method comprises the steps of obtaining a to-be-restored image of a blind person face; inputting the to-be-restored image into the image restoration model to obtain a target restoration image output by the image restoration model; wherein the image restoration model comprises a degradation prediction network, a text prompt module, a visual prompt module and a decoding module; according to the method, a degradation prediction network can effectively evaluate the degradation degree of an input to-be-restored image and provide a corresponding degradation probability weight so as to enhance the prompt generation capability of a subsequent task; moreover, the introduction of the degradation prediction network improves the adaptability of the image restoration model, and enables the model to adjust a prompt strategy according to the degradation degree, thereby optimizing the final image restoration effect of the face of the blind person, and improving the fidelity of image restoration.

The invention belongs to the technical field of rice panicle shielding image data processing, and provides a rice panicle shielding image restoration method and system based on a generative adversarial network, and the method comprises the steps: collecting rice panicle images at different angles and under different illumination conditions to construct a data set, employing a target detection model to carry out the positioning of a rice panicle region, and classifying the shielding types, and extracting a visible area of the rice spike by adopting a semantic segmentation network, and repairing a sheltered area under a generative adversarial network framework to obtain a complete rice spike image. The method is suitable for complex scenes such as leaf shielding, inter-panicle mutual shielding and mixed shielding, texture details and structure consistency of the repaired image can be guaranteed, acquisition of complete phenotype information of the rice panicles is achieved, and reliable data support is provided for rice yield estimation and precision agricultural management.

The invention relates to the technical field of image processing and cultural relic protection, in particular to a cultural relic image virtual restoration method based on texture reconstruction and color correction, and the method comprises the following steps: S1, obtaining a digital image of the surface of a to-be-restored cultural relic, and carrying out the denoising and brightness equalization processing of the digital image; s2, identifying and segmenting a damaged area in the image; s3, selecting an optimal sample block which is most matched with the structure of the region to be filled; s4, performing adaptive affine transformation on the optimal sample block to generate a reconstructed texture block; s5, carrying out linear transformation to obtain a repaired texture block after color correction; and S6, seamlessly embedding the repaired texture block after color correction into the damaged area. According to the method, the texture direction alignment and the color distribution correction are combined, so that the consistency of the damaged area and the surrounding image in structure and color is realized, and the naturalness and integrity of cultural relic image restoration are remarkably improved.

The application discloses a kind of detection and defense method of counterpatch, respectively based on abnormal positioning and edge detection.The detection method of counterpatch based on abnormal positioning utilizes clean image to train the generator-adversary network of encoder-decoder structure;The output image is obtained by inputting the image to be detected into the generator-adversary network, and the absolute error is obtained by subtracting and taking absolute value;The image region whose absolute error is greater than error threshold is the region where counterpatch is located;The detection method of counterpatch based on edge detection converts the image to be detected into gray scale image and carries out edge detection, and obtains edge image;The edge lines in edge image are connected into a closed area one by one, and the closed area whose area is less than the area of counterpatch region is the region where counterpatch is located.The application can detect counterpatch based on the two schemes of abnormal positioning and edge detection respectively, and blacken the region or use image restoration algorithm to restore the region to defend counterpatch.

The invention provides an image restoration model training method, an image restoration method and related equipment, and relates to the technical field of image processing. The training method is applied to a cloud server, and comprises the following steps: acquiring a training sample set comprising a plurality of high-definition sample images; performing compression processing on the high-definition sample image to obtain a corresponding sample compression image; determining a sample residual image according to the high-definition sample image and the sample compressed image; based on an image restoration model, performing restoration processing on the sample compressed image according to the sample residual image to obtain a prediction restoration image; determining target loss according to the prediction restoration image, the high-definition sample image and the sample residual image; and training an image restoration model according to the target loss to obtain a trained image restoration model. The image restoration model trained through the method can effectively restore the quality degradation problem caused by loss of compression information, so that the restoration precision and the playing image quality can be improved when the terminal equipment deployment model carries out image restoration.

The invention relates to a document image restoration method and system, and the method comprises the steps: obtaining a to-be-restored document image, carrying out the preliminary reconstruction through a pre-trained preliminary restoration network, and obtaining a preliminary restoration image; document structured information of the to-be-repaired document graph is extracted, and text features are obtained based on the document structured information; performing text content coding, position information coding and confidence coefficient coding on each text feature, and performing fusion through a feature fusion module to generate a text control feature; inputting the text control feature and the preliminary restoration image into a ControlNet together to generate a control signal; and inputting the control signal into a pre-trained potential diffusion model to de-noise the potential spatial features step by step, and decoding the de-noised potential spatial features by using a pre-trained potential decoder to generate a repaired document image. The method and the device have the effect of improving the document image restoration efficiency and the restoration effect.

The invention relates to the technical field of image restoration, in particular to an image restoration system combining an attention mechanism and a generative adversarial network. The input unit receives an input image and a mask image and preprocesses the input image and the mask image; the generation unit generates a repaired image through a coding layer, a double-flow parallel attention bottleneck layer and a decoding layer, the double-flow parallel attention bottleneck layer generates global structure features by using a global context attention module, a mask-guided sparse attention module extracts local texture features, and the global structure features and the local texture features are fused through a self-adaptive gating fusion module; the discrimination unit adopts a multi-scale structure, carries out authenticity judgment on images with different resolutions through parallel sub discriminators, and introduces a gradient map as an additional channel; the training unit optimizes the generative adversarial network by using a composite loss function, wherein the composite loss function comprises adversarial loss, reconstruction loss and multistage frequency loss; the output unit outputs a final restored image; according to the system, the structure consistency and texture fidelity of image restoration are effectively improved.

The invention discloses a power transmission line defect detection method based on power transmission line defect image augmentation, and the method comprises the steps: firstly obtaining a normal image and a defect image containing insulator sheets in a power transmission line, and extracting a mask region of each insulator sheet in the normal image through employing a segmentation cutting model; and inputting the region as a to-be-restored region into the image restoration model, and generating a simulated defect image of the missing insulator sheet. Semantic similarity evaluation is carried out on the simulation images through a screening module based on perceptual similarity, and high-quality defect samples are screened to construct a defect image sample set. The sample set is used for training a target detection model based on a YOLOv8x structure, and a random cutting strategy of a target area protection mechanism is introduced in the training process to enhance the generalization ability of the model. And finally, the insulator defect in the to-be-detected image is automatically identified by using the trained detection model.

The invention discloses a video restoration method, a video restoration model and a video restoration model construction method. The method comprises the following steps: multiplexing complete parameters and architecture of a pre-training image restoration model, and carrying out unified time sequence expansion processing on the model: adding a time sequence convolution layer and an attention layer in an image encoder and a decoder to construct a time sequence encoder and a time sequence decoder; integrating a time sequence attention module and position embedding in the image restoration network to construct a video restoration network; and finally, combining the expanded components and inheriting pre-training parameters to complete low-cost construction of the high-performance video repair model. When the model is applied, time-space features are extracted through time sequence coding, global time-space cooperative enhancement is carried out through a video restoration network, and then a high-definition video is reconstructed through a time sequence decoder. The problems of inter-frame hopping, detail smearing and high training cost during video restoration are solved, and the high-quality restored video with rich details and coherent time sequence is output while the training resource consumption is reduced.

The invention relates to the technical field of image processing, in particular to an online calibration method and system for hydrological flow data, and solves the technical problem of low flow calculation precision caused by rough window segmentation and illumination interference in the prior art. The method comprises the following steps: collecting a water surface video in a river water flow process, and setting a velocity measurement line area in the water surface video along a water flow direction; according to the gray level distribution of the velocity measurement line area, carrying out illumination state judgment, and according to a judgment result, carrying out image restoration on an illumination abnormal area in the water surface video to obtain a restored water surface video; constructing a space-time image for the repaired water surface video, performing corner detection, clustering corners to divide a plurality of windows, and constructing a window weight for each window; and weighting the texture angle of each window according to the window weight of each window to obtain a weighted texture angle, and calculating hydrological flow data according to the weighted texture angle.

The invention discloses a defense method for resisting an infrared detection confrontation sample attack, and belongs to the technical field of calculation, reckoning or counting. The method comprises the following steps: firstly, partitioning an input infrared image according to a fixed size, executing two-dimensional discrete cosine transform, calculating the mean square difference of the image on each color channel before / after recompression, then converging into a frequency spectrum heat map, and binarizing to obtain a rough mask for positioning an adversarial patch; inputting the original image into the image segmentation model, calculating the intersection-union ratio of the segmentation mask and the rough mask, removing redundancy, and fusing to generate a final patch area mask; then, according to the patch area mask, original image pixels are removed, and a general image completion algorithm is called to recover a removed area; and finally, sending the complemented image into a deep learning target detector to realize robust detection under the infrared confrontation sample attack. According to the method, defense is carried out on adversarial sample attacks under infrared detection for the first time, and adversarial samples are effectively defended by adopting adversarial sample positioning, image segmentation and image restoration.

The invention provides an image restoration method and device, equipment, a storage medium and a program product, and relates to the technical field of image processing. The image restoration method comprises the following steps: acquiring an image analysis result of a to-be-restored image; based on the image analysis result and a preset experience knowledge base, an image restoration scheme of the to-be-restored image is determined, and the image restoration scheme comprises at least one image restoration operation and an execution sequence of the image restoration operations; and based on the image restoration scheme, executing each image restoration operation according to the execution order to restore the to-be-restored image.

Embodiments of the invention disclose an image processing method and apparatus, and an electronic device. The method comprises the steps of obtaining a first image block including a mirror surface area from a first panoramic image; determining a corresponding area of the to-be-removed object in the first image block; removing the to-be-removed object from the first image block based on the corresponding area of the to-be-removed object by using an image restoration model, and performing image restoration on the first image block after the to-be-removed object is removed to obtain a second image block; and replacing the first image block in the first panoramic image with the second image block to obtain the second panoramic image, so that the to-be-removed object in the panoramic image can be removed, the condition of mistakenly removing the object can be avoided, and seamless image reconstruction can be realized through accurate image restoration.

The invention provides an integrated image restoration method and system, and belongs to the field of image processing. The method comprises the steps that an integrated image restoration model TGMIR based on text guidance is constructed, and the integrated image restoration model TGMIR is used for mapping a text prompt to a semantic space consistent with image features and achieving hierarchical semantic regulation and control on three levels of channel attention, space attention and cross-modal collaborative attention; and recovering the image by using the integrated image recovery model TGMIR. The invention aims to break through the core limitation of the existing integrated image restoration model under the multi-degradation condition, including the problems of insufficient degradation semantics, obvious cross-degradation interference, weak modal cooperation capability, insufficient feature fusion and the like.

The invention discloses an image super-resolution reconstruction method based on frequency-space double-domain feature fusion and contrast enhancement, and belongs to the technical field of image processing. The objective of the invention is to solve the technical problems of limited receptive field, high-frequency detail loss and unstable training of an existing super-division model. The core of the technical scheme is that after a low-resolution image is preprocessed, the low-resolution image is input into an improved Transform backbone network, a Fourier enhancement lightweight module (FEL) is inserted into every three residual Transform blocks in a depth feature extraction stage, and a global structure and local textures are captured in a frequency-space double-flow parallel mode; after features are fused through long jump connection, a contrast perception channel attention module (CAA) is input to enhance detail response; layerNorm4D is adopted in the whole process to achieve normalized alignment, and finally a high-resolution image is output through up-sampling. The method is high in global feature capturing capacity, high-frequency details are completely reserved, small sample scene training stability is remarkably improved, and the method can be widely applied to scenes such as image restoration and monitoring image quality improvement and has high practical value.

The invention relates to a multi-view consistent object removal method for a three-dimensional scene based on text guidance. The method comprises the following steps: generating an effective bounding box based on a text and a target detection model; generating a multi-view segmentation mask by using a segmentation model SAM; generating a restored RGB image and a restored depth map based on the image restoration model and the depth estimation model; respectively extracting image and depth features based on the image and a depth encoder, and carrying out spatial alignment and fusion; and constructing a multi-loss function to optimize 3D Gaussian scene parameters. The method fully extracts text, image and depth information, realizes effective utilization of multi-modal information, ensures cross-view consistency in combination with multi-modal contrast learning, realizes accurate and efficient object removal in a large-scale 3D Gaussian scene, and shows significant advantages in text-guided 3D editing and cross-view consistency maintenance.

The present disclosure relates to systems, methods, and non-transitory computer-readable media that modify digital images via scene-based editing using image understanding facilitated by artificial intelligence. For instance, in one or more embodiments, the disclosed systems generate, utilizing a segmentation neural network and without user input, object masks for objects in a digital image. The disclosed systems determine foreground and background abutting an object mask. The disclosed systems generate an expanded object mask by expanding the object mask into the foreground abutting the object mask by a first amount and expanding the object mask into the background abutting the object mask by a second amount that differs from the first amount. The disclosed systems inpaint a hole corresponding to the expanded object mask utilizing an inpainting neural network.

The invention relates to the technical field of intelligent models, and discloses an image generation method and device, computer equipment, a storage medium and a program product. Similar preset prompt words, a potential classification list of a to-be-generated image and confusion category pairs are determined on the basis of original prompt words, so that the confusion category pairs are utilized to carry out classification correction; misjudgment of the subject category of the to-be-generated image is avoided, and the accuracy of the subject category is improved. And if the theme category is a character theme, optimizing the original cue word through the scene type of the to-be-generated image to obtain a target cue word, so that the image generation model performs image generation by using the target cue word with rich details and the scene type. When iteration of the image generation model is not finished, face restoration is carried out on the intermediate image generated by the image generation model, so that generation time prolonging caused by image restoration after iteration is finished is avoided, image generation efficiency is improved, a face collapse phenomenon is avoided, and image generation quality is improved.

A system or method for inpainting can be aided through the use of machine learning and ground truth data training. The training of machine-learning inpainting models through the use of ground truth image data may add efficiency and precision to the field of image inpainting. Furthermore, machine-learning inpainting models can aid in the non-deterministic prediction of a variety of data types and can be applicable to the removing and / or replacing of a variety of data types. The trained models can be enabled to make predictions without ground truth reassurance due to calibrated parameters tuned through the training.

The invention discloses a medical image restoration method based on text-driven prompting and double-domain modeling, and the method comprises the steps: S10, inputting an image ILQ, and extracting a shallow feature Fs through a convolution layer; s20, enabling the Fs to generate a depth feature Fd by using an N-level encoder decoder network with jump connection; on a decoder side, each level comprises NT dual-domain converters and an up-sampling convolution layer; each decoding hierarchy is guided by a text-driven prompt module enhanced by LLM, and the text-driven prompt module provides related task information and generates information prompt to guide space and frequency modeling in the two-domain converter; and S30, processing the depth feature Fd by using a convolutional layer, predicting a residual image Ir, and adding the residual image Ir to an input ILQ image to obtain a restored image. The method supports a plurality of medical image recognition tasks with different degradation types and severity, guides the generation of high-quality task related prompts, and supports the realization of more efficient and general image restoration in a plurality of degradation scenes.

The invention relates to an abnormal phased array radar echo image restoration method based on a generative adversarial network, and the method mainly comprises the following steps: carrying out the preprocessing of radar echo image data, and generating a missing region mask; then, constructing a generative adversarial network model based on a double-flow encoder-decoder, extracting features of input image data and masks through a double-flow encoder, and performing feature fusion; then, a strong echo color correction module and a local detail enhancement module are utilized to perform enhancement processing on the fusion feature map, and detail textures of a shielding region and a strong echo region are restored; reconstructing the image layer by layer through a double-current decoder, and generating a repaired complete image; and finally, carrying out adversarial training by adopting a double-flow discriminator, optimizing a generator, and ensuring the authenticity and accuracy of the repaired image. According to the method, the abnormal region, especially the strong echo region, in the radar echo image can be effectively restored, and the image restoration precision and efficiency are improved.

The invention discloses a high-precision image restoration method, which comprises the following steps: acquiring a reference image and a to-be-restored image, calculating initial affine transformation, and generating a coarse alignment reference image; extracting a reference feature of the coarse alignment reference image and a target feature of the to-be-restored image; modeling boundary uncertainty of the to-be-restored image to obtain boundary uncertainty information; executing bidirectional feature mapping and consistency verification between the reference feature and the target feature to obtain a consistency verification result; performing semantic-guided local adaptive transformation to generate local adaptive transformation information; collaboratively fusing the reference features and the target features based on three items of information to generate fused features; and reconstructing the to-be-restored image by using the fusion features. According to the method, through multi-module cooperation of boundary probability modeling, bidirectional consistency verification and semantic adaptive transformation, the problems of stiff boundary processing, feature error propagation and local transformation mismatch in the prior art are solved, and the structural consistency and detail fidelity of the repaired image are improved.

The invention discloses a dynamic shelter restoration method and system based on continuous streetscape panoramic images, and the method comprises the steps: firstly obtaining a to-be-restored target panoramic image A and a to-be-restored reference panoramic image B, and generating an original pixel-level shelter mask; secondly, extracting matching points among the panoramic images, realizing cross-view geometric alignment of the images, and obtaining a target perspective view C and a reference perspective view D through perspective re-projection; and then inputting the target perspective view C and the reference perspective view D into a three-dimensional reconstruction framework, and performing three-dimensional point cloud reimaging under the camera pose of the target perspective view C by using a depth inspection mechanism. And finally, carrying out image restoration on a three-dimensional point cloud re-imaging result, restoring to a panoramic coordinate system, splicing with an original panoramic image, and outputting a shielding-free panoramic image. According to the method, it is ensured that the repairing result conforms to the authenticity and geometric consistency of the geographic space, and the problems of overlapping conflicts and visual tearing during multi-view projection fusion are effectively solved.

The invention discloses a blurred bar code image processing method and system fused with super-resolution restoration, and relates to the technical field of bar code image restoration, and the method comprises the four steps: firstly, extracting the frequency domain and texture complexity index of a target blurred bar code image, and achieving the automatic classification of dynamic, defocusing and Gaussian blur, and exclusive denoising; building a double-branch feature extraction network, extracting a structure and texture feature map, and generating a cross-scale fusion feature map; outputting a bar code image restoration first draft; and finally, constructing a dual-objective loss function for iterative optimization, and outputting a final clear image. The beneficial effects of the method are that multi-dimensional feature classification enables fuzzy type identification to be accurate, and exclusive denoising lays a high-quality foundation for subsequent restoration; the multi-modal feature fusion is combined with the layered super-resolution technology, a double-target loss function and a closed-loop verification mechanism ensure that the final image can be stably decoded, and the method is suitable for mainstream barcode types and has high practical value in industrial barcode recognition scenes.

A device comprising an image processor configured to implement: a first machine learning model for performing restoration processing on degraded image data; and a second machine learning model for recognizing areas of an image requiring processing emphasis during the restoration processing, wherein the output of the second machine learning model is an input to the first machine learning model to optimize the restoration processing.

The invention provides an image restoration method and system based on reverse cascade energy structure co-evolution, a medium and equipment, and belongs to the technical field of image processing and image restoration. The system comprises: a front-end CNN configured to perform visual feature extraction to obtain an initial feature map; the distance map generator is used for calculating the shortest Euclidean distance from each pixel in the damaged area to the known area to obtain a distance map; the damaged area is divided into equidistant concentric annular belts; the boundary statistical encoder is used for calculating a statistical moment of each annular belt and a joint feature of each pixel in each annular belt; the reverse band cascade HDNN is used for calculating the Hamiltonian amount of each annular band by using a Hamiltonian function and executing half-step symplectic evolution according to a sequence from an outer band to an inner band, and the Transform repair network is used for repairing a damaged area by using the Transform repair network by taking an evolution result as input and outputting a repaired image. According to the scheme, the structural consistency of damaged image restoration can be ensured, distortion is avoided, and the restoration effect is improved.

The invention discloses an image restoration method, system and device, a medium and a product, and belongs to the field of power grids, and the method comprises the steps: obtaining a to-be-restored image of a power scene under an extreme weather condition, inputting the to-be-restored image to an image restoration model, and obtaining a target image, the image restoration model is obtained by guiding an initial restoration model under a semi-supervised framework through an evaluation model to carry out iterative training, in each iteration, the label-free data is input into the first restoration model to obtain a restoration result, the restoration result is input into the evaluation model to carry out quality evaluation to obtain an evaluation signal, and the evaluation signal is sent to the semi-supervised framework; dynamically adjusting model parameters of a current second repair model based on the evaluation signal to guide the current second repair model to learn key features of the power equipment under extreme weather conditions, the evaluation model being obtained by performing all-parameter supervision fine tuning training on a multi-modal large language model based on power scene data; therefore, the generalization ability and the repairing effect of the repairing model can be improved by implementing the method and the device.

A system and a method for medical image processing are provided. The method includes: obtaining a first medical image; obtaining a trained image perception restoration model; the trained image perception restoration model includes an image quality perception sub-model and an image restoration sub-model; and inputting the first medical image into the trained image perception restoration model to obtain a second medical image. The image quality perception sub-model is configured to determine either or both of a first quality evaluation value of the first medical image and a second quality evaluation value of the second medical image, the image restoration sub-model is configured to determine the second medical image, and the quality of the second medical image is higher than that of the first medical image.