133results about How to "Rebuild fast" patented technology

The invention discloses an image super-resolution reconstruction algorithm based on a shallow convolution neural network and a deep convolution neural network. The method comprises the following steps: 1) selecting a training sample and a testing sample; 2) performing extracting, mapping, upsamping and multi-scale conversion to the deep network characteristics; 3) extracting the characteristics of the shallow network; and 4) combining the shallow network and the deep network. Compared with the prior art, with regard to the single image reconstruction and video sequence reconstruction, the algorithm of the invention can accurately and effectively reconstruct the model of a high-resolution image so as to obtain a good reconstruction effect. In addition, multi-scale detail characteristic extraction is achieved, and the algorithm has better results than that of other existing algorithms. The reconstruction speed is also fast.

The invention discloses a mura defect detection method based on sample learning and human visual characteristics, which belongs to the TFT-LCD display defect detection field. According to the invention, the method comprises the following steps: firstly, utilizing the Gaussian filter smoothing and Hough transform rectangle to preprocess the TFT-LCD display image, removing a large amount of noise and segmenting the image areas to be detected; then, using the PCA algorithm to conduct learning to a large amount of defect-free samples; automatically extracting the differential characteristics between the background and the target and re-constructing a background image; and then, thresholding the differential characteristics between a testing image and the background; through the reconstructing of the background and the threshold calculating, jointly creating a model. According to the invention, based on the training sample learning, a relationship model between the background structure information and the threshold value is established; and a self-adaptive segmentation algorithm based on human visual characteristics is proposed. The main purpose of the invention is to detect different mura defects in a TFT-LCD, to raise the qualification rate and to increase accuracy for the detection of mura defects.

The invention provides a remote sensing image super resolution reconstruction method and a system based on a depth convolution network. The method comprises steps: a to-be-processed remote sensing image is converted to YCbCr space from RGB space, and brightness space and chromaticity space are separated; a multilayer depth convolution network is built, a super resolution reconstruction model is built based on the multilayer depth convolution network, the super resolution reconstruction model is used for reconstructing the brightness space, and brightness space after reconstruction is obtained; with the brightness space after reconstruction as a guide graph, the chromaticity space is guided for joint bilateral filtering, and chromaticity space after reconstruction is obtained; the brightness space after reconstruction and the chromaticity space after reconstruction are integrated, the to-be-processed remote sensing image after integration is returned to the RGB space from the YCbCr space, a super resolution image is obtained, and the super resolution image has a higher resolution than the to-be-processed remote sensing image. The above method and the system, in a condition of not relying on a multi-temporal remote sensing image sequence in the same scene, realize super resolution reconstruction in view of the remote sensing image, and the image resolution is enhanced.

The invention discloses a three-dimensional color reconstruction system and method based on a PMD camera and a photographing head. The system comprises an independent calibration module, an integrated calibration module and a three-dimensional reconstruction module, wherein the independent calibration module is used for calibrating the PMD camera and the photographing head to obtain inner parameters, outer parameters and distortion parameters of the PMD camera and the photographing head; the integrated calibration module is used for solving corrected images corresponding to images collected by the PMD camera and the photographing head respectively, and a relative rotation torque and a translation vector between a coordinate system of the PMD camera and a coordinate system of the photographing head according to the inner parameters, the outer parameters and the distortion parameters; the three-dimensional reconstruction module is used for selecting pixel points at four edges in the corrected images, solving corresponding points on the photographing head according to the relative rotation torque and the translation vector to obtain the corresponding image edges of the images collected by the photographing head, cutting the images and storing the cut images as new images. According to the three-dimensional color reconstruction system and method, the three-dimensional color reconstruction is implemented on the basis of the PMD camera and the photographing head.

The invention relates to a foot surface three-dimensional information obtaining method based on multiple RGB-D cameras. According to the method, an all-dimensional measuring system built by the RGB-D cameras is applied to rapid obtaining of foot surface three-dimensional information and rapid reconstruction of a foot surface model, and complete foot curved surface three-dimensional coordinates can be colleted at a time without help of any gauge points. The RGB-D camera all-dimensional measuring system is used for rapidly obtaining depth information of a foot type curved surface at various view angles, the depth information is converted into three-dimensional point clouds, the three-dimensional point clouds, obtained by different RGB-D cameras, of local foot curved surfaces are registered, the overall model of the foot type curved surface is reconstructed, and characteristic parameters of a measured foot type can be extracted according to actual requirements. The foot surface three-dimensional information obtaining method based on the multiple RGB-D cameras is high in reconstruction speed, high in measuring efficiency, easy to implement, low in measuring cost and good in use effect.

The invention discloses a multi-channel magnetic resonance image reconstruction method based on deep learning, and relates to a multi-channel magnetic resonance image reconstruction method. The methodcomprises the following steps: acquiring multi-channel magnetic resonance images, and forming a training set by a sensitivity mapping graph, an under-sampled zero-filled multi-channel magnetic resonance image and a full-sampled synthetic image; establishing a multi-channel deep learning network model for magnetic resonance image reconstruction; constructing a loss function of the network; training multi-channel magnetic resonance image reconstruction network model parameters; reconstructing the target under-sampling multi-channel magnetic resonance image; obtaining an end-to-end mapping function from the under-sampled multi-channel image to the complete magnetic resonance image corresponding to the network model and a network model loss function by adopting a residual connection mode forthe iteration block; training magnetic resonance image reconstruction network model parameters of a residual connection mode; and reconstructing the target under-sampling multi-channel magnetic resonance image by using the residual connected network model. The method has the advantages of high reconstruction speed and good reconstruction effect.

In a K-space SENSitivity Encoding (KSENSE) magnetic resonance parallel imaging method the sensitivity distribution of MR reception coils; is calculated and based on the sensitivity of the coils, signals from the respective coil merging channels are merged. The merged data are used to perform k-space data fitting and optimal fitting parameters are found. The fitting parameters are used to remove artifacts in the reconstructed image. The KSENSE method, compared to SENSE, mSENSE and GRAPPA, has the advantages of the image reconstructed by KSENSE having an optimized signal-to-noise ratio (SNR) that is superior to that of an image reconstructed by GRAPPA under the same conditions and approximates that when mSENSE is used, and an image reconstructed by KSENSE has relatively low residual artifacts and an artifact intensity, as a whole, that is superior to that of an image reconstructed by SENSE or the like under the same conditions and equivalent to that when GRAPPA is used, and, compared to GRAPPA that also performs operations in k-space, KSENSE has a higher reconstruction speed.

The invention discloses an infrared temperature measurement imaging device and a detection method thereof. The device comprises an infrared temperature measurement imaging module, a three-dimensional imaging module, a data processing module and a correction processing module. The infrared temperature measurement imaging module is used for acquiring two-dimensional temperature image data of a detected object. The three-dimensional imaging module is used for acquiring position data of a detected object surface. The data processing module is used for acquiring three-dimensional temperature data of the detected object surface according to the position data of the detected object surface, the two-dimensional temperature image data of the detected object and other parameters. The correction processing module is used for correcting the three-dimensional temperature data of the detected object surface according to a position relation of the detected object and the infrared temperature measurement imaging module and acquiring corrected temperature data. The device can acquire three-dimensional temperature distribution of the detected object. Through correcting the temperature of the detected object surface, the accurate temperature of the detected object surface can be acquired.

The invention discloses a virtual view synthesis method based on homographic matrix partition. The virtual view synthesis method based on homographic matrix partition comprises the following steps of 1) calibrating left and right neighboring view cameras to obtain the internal reference matrixes of the left and right neighboring view cameras and a basis matrix between the left and right neighboring view cameras, deriving an essential matrix from the basis matrix, performing singular value decomposition on the essential matrix, and computing the motion parameters including a rotation matrix and a translation matrix between the left and right neighboring view cameras; 2) performing interpolation division on the rotation matrix and the translation matrix to obtain sub homographic matrixes from left and right neighboring views to a middle virtual view; 3) applying the forward mapping technology to map two view images to a middle virtual view image respectively through the sub homographic matrixes, taking the mapping graph of one of the images as a reference coordinate system and performing interpolation fusion on the mapped two images to synthesize a middle virtual view image. The virtual view synthesis method based on the homographic matrix partition has the advantages of being high in synthesis speed, simple and effective in process and high in practical engineering value.

The invention relates to a CT imaging method which is based on multiple scanning and used for CT imaging, namely, super-visual field CT imaging, when the scanning visual field of CT equipment can not cover an object to be detected. The method comprises the following steps: a projection data set is obtained by multiple scanning and collection and the CT image of the object to be scanned is obtained by that a scanning reconstruction algorithm is operated to process the projection data set. The invention is characterized in that when the object to be detected is scanned, the area covered by each scanning is allowed to have overlapped parts and the scanning reconstruction algorithm used for processing the projection data set does not need to rearrange the data. The invention has the main advantages that the size of the object to be detected is not limited by the CT scanning visual field, the scanning mode is flexible; as the reconstruction algorithm does not need to rearrange the data, interpolation calculation needed by the rearrangement and the degrading of the reconstructed image caused by the rearrangement are avoided, therefore, the reconstruction speed is fast and the resolution of the reconstructed image is high.

The invention discloses a laser and image data fused three-dimensional reconstruction method. The method comprises the steps of obtaining three-dimensional point cloud data of a laser radar and RGB image data of a camera in a to-be-reconstructed three-dimensional scene; mapping the three-dimensional point cloud data of the laser radar to a two-dimensional image space of a camera, and constructinga sparse depth map; converting the sparse depth map into a dense depth map, and obtaining RGB information and depth information of each frame of image; constructing a crystal grid of a to-be-reconstructed three-dimensional scene, mapping the depth information of each frame of image into the crystal grid, and determining the surface of a reconstructed three-dimensional model; and mapping the RGB information of each frame of image to the surface of the reconstructed three-dimensional model to form texture information, and obtaining the reconstructed three-dimensional model. The invention furtherdiscloses a laser and image data fused three-dimensional reconstruction system. Data generated by passive three-dimensional reconstruction and data generated by active three-dimensional reconstruction are fused and intercommunicated, and real-time reconstruction of a three-dimensional scene is achieved.

The invention provides a measurement method for the pose of a wind tunnel model combining stereoscopic vision with a gyroscope, belongs to the field of computer vision, and relates to a high-precision measurement method for the pose of a thrown object model in a complex wind tunnel environment. A mode that stereoscopic vision is combined with the gyroscope is adopted in the measurement method to achieve high-precision measurement of the pose of the thrown object model in the wind tunnel environment. The gyroscope is adopted to measure information of rolling angles of the model, binary rolling angle values are directly displayed by means of coded mark points, and sequence image acquisition is carried out on the coded mark points of the thrown object model in a movement field through a left high-speed camera and a right high-speed camera; after picture processing is conducted, pitching, rolling and position information of the thrown object model is obtained according to position information of the coded mark points, and rolling parameters of the model are obtained through binary decoding of the coded mark points. The method is simple and reliable, the problem that measured data are difficult to transmit in the wind tunnel environment is effectively solved, influences of noise are effectively suppressed through the high-brightness mark points, image acquisition quality is improved, and pose measurement precision is improved.

The invention relates to a three dimensional reconstruction method of human body surface under a complex scene. The invention automatically extracts the three dimensional information of the human bodysurface on the basis of robust image foreground extraction and information fusion theory. Firstly, the foreground image of a human body target is extracted from a multiple-view image collected by each camera, the probability of the foreground and background to which each point belongs is calculated to obtain the foreground similarity probability graph; then, the information of each view is fusedto calculate the occupation probability of space voxel by a human body, so that the three dimensional surface of the human body is reconstructed. The invention has the advantages of easy realization,strong robustness, high precision, high reconstruction speed and the like and can provide quick and reliable three dimensional surface of the human body in relevant application fields, such as man-machine interaction, virtual reality, human body information digitization, human body kinematic analysis and the like.

The invention discloses a method for SPEED rapid magnetic resonance imaging based on wavelet domain sparse representation. The method for SPEED rapid magnetic resonance imaging based on wavelet domain sparse representation comprises two steps of regular under-sampled data collection and SPEED image reconstruction based on wavelet domain sparse representation. The regular under-sampled data collection comprises two steps of k-space regular under-sampling and k-space central domain complete sampling. The SPEED image reconstruction based on wavelet domain sparse representation comprises seven steps of zero filling Fourier reconstruction, discrete wavelet transformation, wavelet domain double-layer sparse ghost model construction, overlapped ghost separation based on a least square error method, discrete wavelet reverse transformation, rectification summation of a plurality of ghost images and image reconstruction. The discrete wavelet transformation is used for SPEED data sparse representation, the sparseness of the image is improved through discrete wavelet transformation, and further the quality of the SPEED reconstruction image is improved.

The invention discloses a magnetic resonance spectrum reconstruction method based on deep learning, and relates to a magnetic resonance spectrum reconstruction method. The method includes: Generatinga time domain signal of the magnetic resonance spectrum by utilizing an exponential function; Establishing a training set of the under-sampling time domain signals and the full-sampling spectrum; Designing a convolutional neural network in the data verification convolutional neural network structure; Designing a bottleneck layer in the data verification convolutional neural network structure; Designing a data verification layer in the data verification convolutional neural network structure; Designing a feedback function in a data verification convolutional neural network structure; Establishing a data verification convolutional neural network structure as a spectrum reconstruction model; Training network optimization parameters; Reconstructing an under-sampling magnetic resonance time domain signal of the target; When undersampling operation is carried out in the time-frequency domain, utilizing the strong fitting capability of the convolutional neural network and the data verification capability of the data verification layer to complete rapid and high-quality reconstruction of the undersampling magnetic resonance spectrum signal.

The invention discloses a magnetic resonance image reconstruction method based on an adaptive orthogonal basis, and the method comprises the following steps: (1) obtaining original k-space data, carrying out the Fourier inversion of the sampled k-space data, and obtaining an original reconstruction image; (2) building a reconstruction model based on compression sensing under the constraint of the orthogonal basis; (3) partitioning the original reconstruction image, randomly extracting some image blocks for the learning of the orthogonal basis, and obtaining a group of adaptive orthogonal basis; (4) employing a hard threshold method to achieve the sparse representation of all image blocks based on the adaptive orthogonal basis; (5) employing a least square method to updating the reconstruction image, and obtaining a current reconstruction image; (6) judging whether the current reconstruction image meets the condition of convergence or not: if the current reconstruction image meets the condition of convergence, the current reconstruction image serves as a final reconstruction image; or else, going to step (7); (7) reducing the value of a regularization parameter, the current reconstruction image serves as the original reconstruction image, and returning to step (3). The method is high in reconstruction speed, and is good in image quality.

The invention discloses a passive single pixel telescope imaging system, and belongs to an optical imaging field. The passive single pixel telescope imaging system comprises an imaging lens, a spatial light modulator, a light beam convergent lens and a data acquisition card, which are orderly arranged on a light path; the spatial light modulator is arranged at an imaging position of the imaging lens; the light beam convergent lens is located on a light path of an output beam of the spatial light modulator; the data acquisition card is provided with a photovoltaic conversion chip; the data acquisition card and the spatial light modulator are connected with a processor through a data line. The imaging system is simple in structure and low in environmental requirement; besides, the acquisition time can be effectively reduced and the recovery of target images under a low sampling rate can be realized. The invention further discloses a passive single pixel telescope imaging method. The imaging method is simple in recovery algorithm and fast in image reconstruction speed, and can shoot and reconstruct the images of large sizes, so that the imaging method has a higher application value in the macro fields like military, remote sensing satellite, astronomy and engineering detection, etc.

The invention relates to an unmanned aerial vehicle navigation map construction system based on an image three-dimensional reconstruction technology. The unmanned aerial vehicle navigation map construction system comprises an unmanned aerial vehicle, a data acquisition assembly and a three-dimensional navigation map construction system, wherein the three-dimensional navigation map construction system comprises an image set input system, a feature point extraction system, a sparse three-dimensional point cloud reconstruction system, a dense three-dimensional point cloud reconstruction system, apoint cloud model optimization system and a three-dimensional navigation map reconstruction system. A scene image set is input into a three-dimensional navigation map construction system, feature point detection is carried out on all images, and a sparse point cloud model of a scene and a dense point cloud model of the scene are reconstructed, and then miscellaneous points of the models are removed, and surface reconstruction is carried out, and optimization is carried out, and a three-dimensional navigation map of the scene is reconstructed. The invention further discloses an unmanned aerialvehicle navigation map construction method based on the image three-dimensional reconstruction technology. According to the unmanned aerial vehicle navigation map construction method, the three-dimensional navigation map is constructed by utilizing the system, and the reconstructed three-dimensional navigation map is close to a real scene, and the precision is relatively high.

The invention provides an analytic method for eliminating a metal artifact of a CT image. With the method, the iterative operation can be avoided. The method comprises: a metal image is segmented from an image reconstructed based on projection data on the multi-energy-spectrum-ray condition; a projection image is solved; an approximate value of an energy spectrum function is searched; an expression of an artifact function is obtained and an artifact image is obtained; the artifact image is subtracted from an original image to obtain a corrected image as shown in an attached drawing. The method is characterized in that the band shaped artifact can be eliminated and the structural feature of original information can be kept; and no priori knowledge of the energy spectrum and original projection data are needed according to the correction method and the reconstruction speed is fast.

The invention relates to a method for realizing three-dimensional back projection by a graphics processor, comprising the following steps: storing projection data collected by a CT (Computed Tomography) processor in a host memory; meanwhile, calculating the Hermite coefficient of each pixel point and the sampling value of a spiral weight in a CT image to be reconstructed, i.e. a target image; copying the projection data in the host memory, the Hermite coefficient of each pixel point and the sampling value of the spiral weight into the display cache of the graphics processor; carrying out texture binding to data in the display cache; carrying out three-dimensional back projection in GPU (Graphics Processing Unit) to obtain a CT reconstruction image; and copying the obtained CT reconstruction image into the host memory. By applying the method disclosed by the invention, the three-dimensional back projection is realized in the GPU, and figures are chosen by virtue of a GPU three-dimensional array storage and three-dimensional texture binding technology based on a float4 data type as the basic element of a two-dimensional parameter array, thereby avoiding the problem that the figures are slowly chosen because parameter arrays are excessive so as to cause that GPU cache utilization is low so that operation efficiency is further improved and a satisfied result is obtained.

The invention discloses an efficient reconstruction method in a compression perceptual system, relates to a data processing method which is used for solving the problem that the precision and the speed cannot be improved simultaneously in the existing reconstruction method and comprises the following steps: firstly arranging a measurement value Y0 into a form for easily realizing a reconstruction algorithm, if one-dimensional reconstruction is required, ensuring that the measurement value is not arranged, and if two-dimensional reconstruction is required, vectorizing the measurement value to obtain Y; then enabling the k to be equal to 1, uk to be equal to 0 and vk to be equal to 0, thus obtaining expressions as shown in the specification; performing non-contribution iteration in an iterative step to compute the times s of the non-contribution iteration; if the change times s of the vk just enable the uk+1 to change, containing the following iterative formulas as shown in the specification in the iterative steps to judge an expression as shown in the specification, then judging whether the judgment is met so as to determine whether to converge the iteration, iterating until the iteration is converged; and finally, if an one-dimensional signal is required to reconstruct, directly using signal sparseness to represent a reconstruction original signal, and if a two-dimensional signal is required to reconstruct, inversely vectorizing a sparseness coefficient u, and using the sparseness of the image to represent the reconstruction original image. The efficient reconstruction method can be applied to the reconstruction of the one-dimensional or two-dimensional signal in the compression perceptual system.

The invention discloses a magnetic resonance imaging method and device. According to the method, after a first deep neural network is used for reconstructing a first image, a first part of k-spatial data and a second part of k-spatial data extracted from complete k-spatial data are subjected to subtraction operation to obtain residual k-spatial data; then the residual k-spatial data is subjected to image reconstruction to obtain a residual image; and finally, the first image and the residual image are added to obtain a magnetic resonance image. In this way, the first part of the really collected k-spatial data is utilized in the processing carried out after the first image is reconstructed, the processing process of data fidelity terms is carried out outside the DNN reconstruction, and theDNN does not need to consider the processing of the data fidelity terms, so that the complexity and network parameters are lowered and reduced, and the image reconstruction speed is high. In addition, the processing process of the data fidelity terms is completed through sparse constraint reestablishing, without multi-time iteration, and the reconstruction speed can be improved on the premise that the image reconstruction stability is improved.

A three-regular magnetic resonance image reconstruction method based on Split Bregman iteration relates to a magnetic resonance image reconstruction method. The invention aims to solve the problems that magnetic resonance image reconstruction methods in the prior art are high in computation complexity, cannot eliminate aliasing artifacts and Gibbs ringing, and have limited image reconstruction precision. The method comprises the following steps: k space data without being sampled are obtained through measurement; regularization term constraint is conducted by using a total variation, short support wavelet and a high order, high vanishing moment wavelet, and a reconstructed image is obtained after Split Bregman iteration; the reconstructed image obtained in the second step is subjected to error determination, and if the error does not satisfy a preset condition, the second step is repeated until the preset condition is satisfied, and the reconstructed image is obtained. The method improves the quality of the reconstructed image while ensuring quick reconstruction.

The invention provides an air-space-reservoir-bank integrated emergency survey data fusion and integrated display system and method. The air-space-reservoir-bank integrated emergency survey data fusion and integrated display system comprises a data acquisition and transmission module, a data extraction module, a data management module, a data registration module, a data splicing module, a data analysis and processing module, a data fusion module and a three-dimensional scene reconstruction module, wherein the data extraction module is connected with the data acquisition and transmission module; the data management module is connected with the data extraction module; the data registration module is connected with the data management module; the data splicing module is connected with the data management module; the data analysis and processing module is connected with the data management module; the data fusion module is connected with the data analysis and processing module; and the three-dimensional scene reconstruction module is connected with the data analysis and processing module. According to the air-space-reservoir-bank integrated emergency survey data fusion and integrated display system, the range and the condition of the affected area can be rapidly and accurately surveyed in detail, and the survey result is obtained.

The invention discloses a cone beam CT (Computed Tomography) local scanning imagery for fluid in pipelines, and relates to a nondestructive examination technology for pipelines in use. The method comprises the following steps: mounting a radiation source and an area-array detector on a round track coiled on a pipe wall before scanning, while scanning, making the radiation source and the area-array detector perform circular motion along the track around the pipeline during scanning, simultaneously, by flowing of fluid, obtaining equivalent spiral cone beam projection data of inclusions-containing fluid, and then, obtaining a three-dimensional image of the fluid inclusions in the to-be-detected pipeline by means of combining a spiral cone beam FDK (Feldkamp) image reconstruction algorithm with a local filter. Mechanical motion of such scanning mode is simple and easy to implement, and can easily adapt to field conditions of pipeline field in-use detection; local reconstruction image of the inclusions-containing fluid in the pipeline has good quality and high resolution.

The invention discloses a dynamic three-dimensional figure reconstruction and live broadcast method based on an RGBD camera, and the method comprises the steps: S1, carrying out the real-time image matting of a collected RGBD image, and carrying out the matting of a figure part of the RGBD image; S2, compressing the RGBD image after image matting, and carrying out real-time stream pushing by adopting a video stream pushing protocol; s3, receiving an image matting result of the RGBD image at a mobile terminal, decompressing the image matting result to generate three-dimensional point clouds, and connecting the point clouds into a plane to obtain a three-dimensional model of the human body; and S4, carrying out binocular rendering on the three-dimensional model at the mobile terminal. An image with a three-dimensional effect and an AR effect can be seen by wearing the three-dimensional display device. The method is relatively small in calculated amount, can be used for reconstruction ata mobile terminal, and is high in reconstruction speed and good in effect. The method is wide in application scene, not only can be applied to real-time three-dimensional character live broadcast, butalso can be applied to multiple fields such as holographic communication and holographic conferences, and a user can watch through AR, VR and holographic equipment.

The invention relates to a large diameter pipeline wall outside CT local scanning imaging method, and belongs to the technical field of CT scanning imaging. The method comprises the following steps: a radiation source and a detector are arranged in a circular track surrounding the center of a pipeline to be detected, and the detector is arranged in an offset manner; the radiation source and the detector circularly move along the circular track to scan in order to obtain the projection data of the annular region outside the pipeline to be detected; and a TVM-POCS reconstruction algorithm and a region scale fitting segmentation technology are combined to reconstruct the image of the annular region outside the pipeline according to the projection data. The large diameter pipeline wall outside CT local scanning imaging method has the advantages of simple and easy scanning mode, short scanning time, low radiation dose, satisfactory solving of the cross cutting problem of the projection data, great reduction of the reconstruction artifact, satisfactory solving of the problem of reconstructed image gray scale unevenness caused by ray beam hardening, and good quality of the finally displayed local reconstruction image of the annular region outside the pipeline.

The application discloses a magnetic resonance imaging method and device. According to the method and device, k-space is divided into multiple areas in a phase encoding direction; each area is imparted a different acquisition density during data acquisition; therefore, variable density acquisition can be implemented for the k-space; the same area may be imparted the uniform acquisition density sothat k-space data of the same area can be subjected to areal image reconstruction by means of conventional parallel reconstruction; a final magnetic resonance image can be formed by using areal imagesof the multiple areas. Therefore, in the application, total image reconstruction time is equal to a product of multiplying the number of the areas by the time of the conventional parallel reconstruction. Compared with the prior art where magnetic resonance imaging performs variable density acquisition, the magnetic resonance imaging method and device have short total image reconstruction time andhigh reconstruction speed and accordingly can provide quick image reconstruction for k-space data under variable density acquisition.

The invention discloses a 4D-CBCT imaging method based on motion compensation, and belongs to the field of computed tomography. The method comprises the following steps: firstly, acquiring high-quality 4D-CBCT data of a patient, and dividing the high-quality 4D-CBCT data into samples and label data; then, requiring a motion compensation learning convolutional neural network of 4D-CBCT data to be constructed, wherein the network is used for establishing mapping between different phase images; secondly, training the network by taking the sample and the label data as inputs to obtain an optimal network parameter weight; and finally, reconstructing 4D-CBCT projection data under clinical scanning with the assistance of the network, and acquiring a high-quality reconstructed image. According tothe method, reconstruction blurring caused by respiratory movement and noise and artifacts caused by data acquisition angle loss can be reduced to a great extent, the scanning period can be shortened.The radiation damage to a subject can be reduced, the quality requirements of clinical analysis and diagnosis can be met, and the tracking efficiency of lung tumors can be improved.