28results about How to "Improve reconstruction quality" patented technology

The invention provides a diffusion-driven channel adaptive point cloud semantic communication method, which belongs to the technical field of wireless communication, and comprises the following steps: obtaining a training point cloud data set, constructing a point cloud feature extraction network based on a hypergraph convolutional neural network as a semantic encoder, and constructing a channel adaptive enhancement module as a channel decoder, a channel adaptive recovery module is constructed at a receiving end as a channel decoder, a diffusion reconstruction network is constructed as a semantic decoder, end-to-end joint training is performed on the point cloud feature extraction network, the channel adaptive enhancement module, the channel adaptive recovery module and the diffusion reconstruction network, the diffusion reconstruction network predicts original point cloud distribution, and the point cloud feature extraction network and the channel adaptive enhancement module are subjected to end-to-end joint training. And a loss function is constructed based on the chamfering distance between the predicted point cloud distribution and the original point cloud. In the communication process, the reconstructed semantic features are input into the diffusion reconstruction network to reconstruct an original point cloud structure. The high-order semantic features of the point cloud can be effectively extracted so as to improve the adaptive capacity of the method to the incomplete point cloud.

The application provides a generative image compression method based on vector quantization, and belongs to the technical field of image and video compression. The method comprises the following steps: obtaining a continuous latent representation of an input image through an analysis transformation module; then performing vector quantization processing to obtain discrete indexes of the continuous latent representation and corresponding quantized features; constructing a continuous index probability distribution; predicting a conditional probability distribution of the discrete indexes through a conditional autoregressive entropy model; calculating a coding rate based on the continuous index probability distribution and the conditional probability distribution; reconstructing an image based on the quantized features and constructing a distortion loss; constructing a rate-distortion loss function and jointly training an image compression model comprising the analysis transformation module, the vector quantization module, the conditional autoregressive entropy model and a synthesis transformation module to obtain a trained image compression model; and compressing the input image by using the trained image compression model. The application can realize rate-distortion joint optimization and collaborative improvement of compression efficiency and reconstruction quality.

The present application relates to the technical field of computer graphics, computer vision and three-dimensional reconstruction, and particularly relates to an immersive scene reconstruction rendering method and system based on adaptive 3D Gaussian sputtering, which acquires scene initial point cloud data and camera parameters, constructs a 3D Gaussian point cloud model containing multiple attributes and generates sky point cloud; based on the mechanism of pulse neurons, the opacity of Gaussian points is adaptively controlled, and according to the gradient information and geometric attributes of Gaussian points, splitting, copying or pruning operations are performed to optimize the distribution of point cloud; through a deep learning network, the camera pose is corrected, large-scale scenes are divided into blocks for parallel training, and a bilateral grid post-processing is combined with multiple loss functions to optimize rendering, and the sky point cloud is subjected to exclusive constraint, so that the technical defects of the prior art are solved, and the immersive scene application requirements in the fields of VR / AR, digital twin and the like are met.

The invention discloses a construction method of a video reconstruction system of a joint information source channel, the video reconstruction system, equipment and a medium. At a transmitting end, the method comprises the following steps: performing multi-frame joint semantic coding on a video sequence, extracting a potential representation simultaneously containing spatial information, time information and high-level semantic information, and directly mapping the potential representation to a wireless channel for transmission; and at a receiving end, a diffusion generation model is constructed based on the denoising network and the diffusion model, and semantic features are introduced to carry out condition guidance on the diffusion denoising process, so that generative reconstruction of the potential representation damaged by noise is realized. According to the method, the diffusion generation model is integrated into a deep joint source channel coding framework, so that the video reconstruction quality and semantic consistency are remarkably improved in a low signal-to-noise ratio and complex channel environment, the cliff effect is effectively weakened, and higher robustness and adaptive ability are achieved.

The application discloses an audio coding model construction method, device and equipment. The method includes the following steps: constructing a network structure of an audio coding model adopting a dynamic frame frequency mechanism; learning network parameters of the model from audio training data; dividing a multi-level residual vector quantizer related to acoustic characteristics in the model into multiple quantizer groups; obtaining audio training data of different frequency bands corresponding to different quantizer groups; and adjusting network parameters of the quantizer groups according to the audio training data of the frequency bands corresponding to the quantizer groups. By using this processing method, different frequency band-specific quantization optimization strategies are adopted for acoustic characteristics of different frequency bands in the case of dynamic frame rate, dynamic frame rate and frequency band reconstruction are coordinated, and low-frequency energy of an audio signal is prevented from covering high-frequency details when the dynamic frame rate is low; therefore, the full-band reconstruction quality is effectively ensured to be uniform, and the audio reconstruction quality is improved.

This invention relates to a hyperspectral snapshot compressed imaging reconstruction method with fused conditional diffusion, belonging to the field of deep learning technology. The method includes the following steps: S1: acquiring a two-dimensional compressed image of the target scene; S2: preprocessing the hyperspectral image training set to generate sample data for training; S3: inputting the preprocessed training samples into a hierarchical conditional diffusion image reconstruction network for training; S4: after the network training is completed, inputting test samples into the network to reconstruct the hyperspectral image and obtain the result. This method, while reconstructing the detailed structure of hyperspectral images, possesses stronger denoising capabilities and has significant advantages over other methods in suppressing artifacts.

The application relates to a method and device for reconstructing an angiography image and an angiography imaging system, wherein the method for reconstructing the angiography image comprises the following steps: acquiring first projection data collected by a target imaging device at a first acquisition angle, and acquiring second projection data collected by an X-ray imaging device at a second acquisition angle; the X-ray imaging device is independent of the target imaging device, and the first acquisition angle is different from the second acquisition angle; reconstructing a four-dimensional angiography image according to the first projection data collected at multiple first acquisition angles and the second projection data collected at multiple second acquisition angles, and obtaining the four-dimensional angiography image. The method can obtain projection data at different angles without changing the collection track of the target imaging device, can reduce the interference of the blood vessel overlapping area, can consider the time resolution, and can improve the reconstruction quality of the 4D DSA reconstruction image.

The invention discloses a spatial-temporal feature enhancement-based dynamic point cloud geometric compression method and related equipment, and the method comprises the steps: carrying out the multi-layer down-sampling of a reference point cloud and an original point cloud, and obtaining L layers of reference compression point clouds and original compression point clouds; entropy coding and decoding are carried out on the L layers of original compressed point clouds to obtain L layers of reconstructed point clouds; l = L; performing up-sampling on the lth-layer reconstruction point cloud to obtain (l-1)-layer extension point cloud, performing geometric coding and decoding on the (l-1)-layer extension point cloud and the (l-1)-layer original compression point cloud to obtain (l-1)-layer coarse reconstruction point cloud; performing time domain modeling based on the (l-1)-layer coarse reconstruction point cloud and the reference compression point cloud to obtain (l-1)-layer fusion alignment features; obtaining (l-1) layers of reconstruction point clouds based on the (l-1) layers of fusion alignment features and the original compression point clouds; performing progressive decrease operation on l, and performing cyclic reconstruction until l is equal to 1; and performing up-sampling according to the first-layer reconstructed point cloud to obtain a final reconstructed point cloud. The method can adapt to complex point cloud compression requirements and can be widely applied to the technical field of point cloud data processing.

The invention relates to a three-dimensional reconstruction method and system for a scene in a three-dimensional Gaussian splash chamber based on plane prior guidance, and belongs to the technical field of three-dimensional reconstruction and computer graphics. The method aims at solving the problems that in indoor scene reconstruction, especially in a weak texture area, geometric reconstruction of an existing method is not accurate, and artifacts are likely to be generated. The method comprises the following steps: firstly, acquiring a multi-view image and camera parameters thereof, predicting depth and normal priori, and extracting a plane region through multi-granularity segmentation and geometric consistency check; a dense point cloud is generated by using depth prior back projection, and a three-dimensional Gaussian splash model is initialized based on a down-sampling point cloud. And finally, under a micro-renderable framework, optimizing the model by combining the planar region geometric constraint, the non-planar region geometric constraint and the global geometric constraint, and completing reconstruction. According to the method, the reconstruction precision of the weak texture plane region is effectively improved, the detail recovery capability of the complex structure region is enhanced, and a reconstruction result with visual high fidelity and geometric accuracy is realized.

This application provides a method, system, electronic device, and medium for high dynamic range (HMR) reconstruction based on a single image. The method includes: processing a high RMR image used for training into a low RMR image to be processed and an optimized low RMR image; using the low RMR image to be processed as input to a partially convolutional U-shaped network in a two-stage deep learning model to train the partially convolutional U-shaped network, thereby obtaining an intermediate low RMR image; using the intermediate low RMR image to train a classical convolutional U-shaped network in the two-stage deep learning model, thereby obtaining a reconstructed HMR image; and inputting a single low RMR image into the trained two-stage deep learning model to obtain the reconstructed HMR image. This application enables HMR reconstruction based on a single image through a two-stage deep learning model, improving the reconstruction quality of HMR images based on a single image.

The invention discloses a frequency domain enhanced single-pixel imaging method and a frequency domain enhanced single-pixel imaging system, which are used for acquiring a single-pixel measurement value corresponding to a target object and performing image reconstruction based on the single-pixel measurement value to obtain a target imaging result. The method comprises the following steps that a single-pixel imaging network is constructed, the single-pixel imaging network comprises an input layer, a plurality of coding layers, a plurality of decoding layers and an output layer which are connected in sequence, the number of the coding layers is the same as that of the decoding layers, and each coding layer comprises a basic convolution layer, a Swindow-Transform module and a frequency domain MLP module which are connected in sequence; based on the single-pixel measurement value, performing iterative training on the single-pixel imaging network according to a double-domain self-supervision method, and obtaining a first reconstructed image and a second reconstructed image in each iteration; and taking the optimal first reconstructed image as a target imaging result. The method can give consideration to both imaging quality and reconstruction speed.

The application provides a 3D Gaussian sputtering method based on deep feature fusion, comprising: extracting multi-view features and monocular depth features through a multi-view Transformer network and a pre-trained monocular depth estimation model respectively; dynamically fusing the two types of features through a content-guided attention module; using a 2D U-Net network to perform depth regression on the fused features to obtain a robust depth distribution; projecting the depth distribution back to a 3D space to obtain Gaussian centers, and then using a 2D U-Net network to predict other Gaussian parameters; and rendering a high-quality three-dimensional model according to all Gaussian parameters. The 3D Gaussian sputtering method combines the complementary advantages of multi-view feature matching and monocular depth prior, significantly improves the reconstruction quality of complex scenes, and solves the problem of low scene reconstruction quality caused by inaccurate matching or missing matching information of traditional feature matching methods in complex scenes.

The application discloses a remote sensing image blind super-resolution method and system based on unsupervised learning and belongs to the technical field of remote sensing image reconstruction. The application is applied to a remote sensing image super-resolution scene in which a degradation model is unknown. First, spatial reconstruction information is transferred from a natural image field to a remote sensing image field by using transfer learning. Then, the spatial-spectral reconstruction information is optimized based on unsupervised learning. By using the transfer learning, the application does not need to be trained on remote sensing images with limited training data, and the advantages of data-driven are explicitly used. By using the unsupervised learning, the application can be generalized to remote sensing image super-resolution reconstruction tasks in various degradation scenes, the spectral reconstruction information is fully learned, and the blind super-resolution precision of the remote sensing image is improved.

The application provides a self-supervised speech denoising method and device, and relates to the field of acoustic signal processing, and comprises the following steps: S1, obtaining a DCUnet network, constructing a speech denoising model by adding a TCM module in the DCUnet network; S2, obtaining a noise speech set, training the speech denoising model through the noise speech set, and obtaining a trained speech denoising model; S3, denoising speech through the trained speech denoising model. The application constructs a speech denoising model based on a deep complex domain DCUnet network, combines a complex domain overall processing strategy, improves the reconstruction quality and denoising performance of the speech signal, dynamically captures the change characteristics of the noise scene through the TCM module in the speech denoising model, enhances the adaptability of the model to dynamic signals in a complex environment, trains the speech denoising model by using an ONT strategy, uses noise speech as training data, does not need clear target speech data, reduces the training data collection cost, and improves the generalization ability of the model.

The present application relates to a genetic algorithm-based optical imaging system sensor optimization configuration method.The present application includes taking the uniformity coefficient of the sensor sensitivity distribution as the optimization target, taking the angle position of the sensor structure light-emitting and light-sensing elements as the optimization object, and obtaining the optimal structure of the optical tomography sensor applicable to the fan-shaped laser through the genetic algorithm.Then, the angles of the light-emitting elements and the light-sensing elements are set according to the optimal sensor structure parameters, so that the optimized sensor is obtained.Finally, the imaging optimization is realized through the imaging system.The present application is not dependent on any specific actual distribution, is applicable to the case without prior information, and has stronger universality.The genetic algorithm is used to realize the optimization of the sensor structure, and the calculation amount of designing an optimal configuration structure with the minimum uniformity coefficient is reduced.

The application relates to a real-world video super-resolution method, system, device and medium, which comprises the following steps: inputting an original video sequence; video embedding: extracting the spatial features of each frame from the original video sequence to obtain first features, and inputting the first features into a double-axis space-time attention mechanism module to obtain second features; wherein the double-axis space-time attention mechanism module comprises a vertical-time attention block and a horizontal-time attention block, the feature blocks generated after the first features are processed through rotation position coding are converted into token sequences, and the token sequences are rearranged and then respectively sent into the vertical-time attention block and the horizontal-time attention block to simulate spatial texture and motion characteristics; space-time reconstruction: time attention is adopted to integrate time information, and a video output with higher space-time quality is generated. Compared with the prior art, the application has the advantages of good video reconstruction quality, high robustness and low cost.

The present application belongs to the field of hyperspectral reconstruction in coded aperture snapshot spectral imaging system, and particularly relates to a hyperspectral image reconstruction method based on spatial-spectral dual-prior frequency domain enhancement, comprising: obtaining a hyperspectral image, a system measurement value of the hyperspectral image and an imaging mask according to a coded aperture snapshot spectral imaging system, inputting the system measurement value and the imaging mask into a trained hyperspectral image reconstruction model to obtain a reconstructed hyperspectral image; the hyperspectral image reconstruction model comprises an imaging mask preprocessing module, a spatial prior denoising reconstruction module and a convolution layer; the spatial prior denoising reconstruction module is adopted to recover spatial information from different angles by using spatial learning and frequency domain learning two branches, the frequency domain learning branch uses the imaging modulation mask as spatial prior to assist in recovering spatial detail features, and the hyperspectral reconstruction quality of the model is improved.

The application discloses a kind of object three-dimensional reconstruction method and device based on deep learning, it is related to three-dimensional reconstruction technical field, wherein, the reconstruction method includes: using camera imaging equipment to collect target object video of target object, and based on target object video, determine target image set, based on target image set, determine target camera parameter and initial camera pose, and correct initial camera pose, obtain target camera pose, input target image set, target camera parameter and target camera pose to preset neural radiance field model, output target radiance field information, based on target radiance field information, construct the object three-dimensional model of target object.The application solves the technical problem that the accuracy of three-dimensional reconstruction of object in related art is low.

The invention particularly relates to a semantic knowledge migration-based high dynamic range imaging method, which comprises the following steps of: acquiring a low dynamic range image, and generating an initial high dynamic range image by utilizing an initial high dynamic range reconstruction model; converting the initial high dynamic range image into a reconstructed image of a standard dynamic range domain through domain migration operation; processing the reconstructed image of the standard dynamic range domain by adopting a semantic segmentation network SAM to obtain semantic priori; processing the reconstructed image of the standard dynamic range domain by combining semantic prior with a semantic prior guidance reconstruction model to obtain a refined reconstructed image; based on a knowledge distillation method, cross-domain semantic priori migration is realized at three levels of content, color and features, and an initial high dynamic range reconstruction model is optimized through a combined loss function; in the reasoning stage, only the optimized initial high dynamic range reconstruction model is adopted to output a final high dynamic range image. According to the method, high-dynamic-range image reconstruction is carried out by effectively utilizing the advanced capability of semantic priori.

This invention proposes a polarization 3D reconstruction method and system based on a prior-guided fusion network. The invention is implemented using a dual-branch architecture and includes a feature correction module that mutually corrects defects in the channel and spatial dimensions. Furthermore, a feature fusion module based on an effective cross-attention mechanism is proposed to fuse polarization and shadow prior features, achieving high-precision surface normal vector estimation and thus reconstructing high-quality 3D targets. Experimental results show that the fusion of polarization and shadow priors significantly improves the reconstruction quality of surface normals, especially for objects or scenes illuminated by complex light sources. In addition, by introducing specular confidence, the angular error of specular reflection regions can be reduced. Finally, because the network can effectively extract and fuse information from different priors, our proposed method outperforms existing deep learning-based polarization 3D reconstruction methods.

The application discloses a photoacoustic tomography sparse reconstruction method based on a diffusion model, and in the photoacoustic tomography reconstruction process, a sparse reconstruction strategy combining a fractional-based diffusion model and a model-based iterative reconstruction method is proposed, a fractional network is used to learn the data distribution of a target image, and the output of the final network is used as prior information of an optimization problem in model iteration to obtain an optimal solution. The method can make the photoacoustic image reconstructed under sparse detection view angles have fewer artifacts than the reconstructed image of the traditional method, and can more effectively and accurately reflect the real information of the target object.

The application belongs to the field of magnetic particle imaging, and particularly relates to an MPI acceleration calibration method based on a system matrix super-resolution network, aiming to solve the problems of long calibration time and poor accuracy of reconstructed MPI images of the existing magnetic particle imaging system matrix reconstruction method. The method comprises the following steps: placing a unit MNP sample in the imaging field of view of a magnetic particle imaging device for sparse calibration scanning, and constructing a system matrix as a low-resolution system matrix; inputting the low-resolution system matrix into a system matrix super-resolution network model after pretreatment, to obtain a super-resolved system matrix row set; performing post-processing on the super-resolved system matrix row set to obtain a high-resolution system matrix; and combining the high-resolution system matrix to solve the magnetic particle concentration distribution of a target object to be imaged, and reconstruct an MPI image of the target object to be imaged. The application can greatly reduce the calibration time while accurately reconstructing the MPI image.

The application discloses a kind of based on rotation matrix's tilt plane imaging method of non-interference coded aperture correlation holography, belong to the field of computational optical imaging and holographic display.Coded phase mask (CPM) is modulated by random or specific design phase to incident light, and the light of each point on the object forms a unique speckle pattern after propagation, without introducing reference light, the light field distribution containing three-dimensional information of object can be recorded;By introducing rotation matrix in CPM, the light wave diffraction between non-parallel planes is simulated, the light field distribution of three-dimensional information of object on the inclined plane is realized, and the limitation that traditional non-interference coded aperture correlation holography system only images in single plane is broken through.The application keeps the advantages of non-interference coded aperture correlation holography technology, such as no interference, simple structure, anti-speckle noise, etc., while significantly enhancing the flexibility and controllability of the system in the selection of imaging plane.The method can be widely applied in three-dimensional holographic display, optical information processing, virtual reality and "digital twin" and other fields, and provides a new technical path for optical imaging and reconstruction in complex scene.

The application discloses a self-adaptive 4D Gaussian splashing high-precision three-dimensional reconstruction system and method, and belongs to the technical field of computer graphics processing. In order to solve the problems of realizing high-precision automatic registration of multi-source heterogeneous data and improving the calculation efficiency, the application comprises the following steps: collecting multi-source heterogeneous data; constructing a multi-modal fusion registration method, including realizing spatial distribution geometric rough registration by combining satellite image data and low-altitude oblique photography data, realizing photometric fine registration by fusing low-altitude laser radar point cloud data and ground acquisition vehicle laser radar point cloud data, and establishing semantic feature auxiliary registration to obtain registered multi-source data; initializing a 4D Gaussian primitive and performing self-adaptive splashing reconstruction to obtain an optimized 4D Gaussian splashing model; designing a cloud-edge collaborative computing architecture for the 4D Gaussian splashing reconstruction, performing distributed parallel processing on the obtained optimized 4D Gaussian splashing model; performing quality evaluation and self-adaptive optimization, and outputting a final self-adaptive 4D Gaussian splashing model.

The application discloses a kind of compressed computing spectral imaging reconstruction method and system based on pulse weighted network, it is related to computing imaging technical field, it can be from compressed measurement value with low energy consumption and high quality Reconstruct HSI data, can greatly improve the energy efficiency ratio of reconstruction algorithm.The scheme is specifically as follows: using coded aperture snapshot spectral imager CASSI obtains the two-dimensional compressed measurement value of known target scene three-dimensional hyperspectral imaging technology HIS data cube.The two-dimensional compressed measurement value obtained is preprocessed, image features are extracted, and the initial value of the reconstruction result of the HSI cube of the known target scene is obtained.The initial value of the reconstruction result is spliced with the coded pattern of the compressed spectral imaging system, and is sent into the trained multi-time-step reconstruction network to obtain the output result of the network, i.e., the reconstruction result of the hyperspectral data of the target scene to be measured.

The application provides a self-adaptive and privacy-protected source channel joint coding online segmentation learning method, comprising the following steps: based on a preset codec network architecture, a segmentation point set containing multiple candidate segmentation points is constructed, each segmentation point divides the encoder into a sending end and a receiving end; based on a context multi-arm tiger machine algorithm, the optimal segmentation point at the current moment is selected from the segmentation point set according to the obtained communication system state information at the current moment; a segmentation reasoning process is performed according to the optimal segmentation point, the sending end obtains intermediate features by processing the input data through the encoder network layer before the optimal segmentation point, and transmits the intermediate features to the receiving end, the receiving end decodes the intermediate features to reconstruct output data; in response to detecting that the data source distribution drifts or the channel condition change amount exceeds a preset threshold, online adaptive segmentation training is triggered. The application realizes the triple balance of reconstruction accuracy, communication overhead and privacy protection.

The method is mainly applied to the technical field of optical information processing and computational imaging. The invention discloses a multi-plane phase hologram generation method based on deep learning and computer equipment, and the method comprises the steps: constructing a holographic generation neural network model, and training the holographic generation neural network model based on the physical constraint condition of an optical system, so as to simulate a light field propagation process; obtaining various structural feature data, and synthesizing the various structural feature data to obtain standard scene data; and inputting the standard scene data into the trained holographic generation neural network model to generate a holographic phase diagram corresponding to the standard scene data, performing light wavefront modulation in the process of simulating light field propagation, and performing reconstruction to obtain a holographic image. According to the invention, the generation efficiency and reconstruction quality of the multi-plane phase hologram can be improved.

The application provides a learning-based tomographic imaging and reconstruction method to measure the scene density distribution in a light reuse manner. During imaging, the light source emits light according to the intensity obtained through pre-learning, the light from different directions reaches the sensor after being absorbed and attenuated by the scene, and the density information of the scene is obtained by calculating and reconstructing the measurement value. The light emission intensity and the reconstruction algorithm are obtained by learning through a neural network. The method models the CT imaging process as a linear fully connected layer, and the weight corresponds to the light emission intensity of the light source during imaging; the reconstruction algorithm is modeled as a nonlinear neural network, which can be optimized according to the characteristics of the scanning geometry. The method requires a small amount of collected data, and the calculation and reconstruction do not require strong prior assumptions, realizes efficient and high-quality CT acquisition and reconstruction, and can be applied to high-speed dynamic scene three-dimensional reconstruction.