63results about How to "Improve reconstructed image quality" patented technology

The invention discloses a fast magnetic resonance imaging method based on residual U-net convolutional neural network and the method comprises three steps of preparing training data, carrying out training based on the residual U-net convolutional neural network, and carrying out image reconstruction based on the residual U-net convolutional neural network. By adding the residual module into the U-net convolutional neural network, the problems of gradient disappearance, overfitting, low convergence speed and the like of the U-net convolutional neural network can be solved, and the quality of rapid MRI imaging based on the U-net convolutional neural network is improved.

The invention discloses a method for three-dimensional representation of stratified structure micrometer CT imaging of a turbine blade heat barrier coating. The method comprises the following steps: preparing a double-layer-structure heat barrier coating micrometer CT scanning sample; firstly, carrying out micrometer CT scanning on the heat barrier coating sample, carrying out ring artifact and beam hardening correction on projection data, then carrying out three-dimensional reconstruction to obtain three-dimensional stratified structure information, establishing a three-dimensional structure analysis model, and carrying out characteristic segmentation and extraction on three components, namely a ceramic heat insulating layer, a bonding layer and matrix alloy; then, analyzing the thickness of the coating, extracting the distribution of holes in the coating, calculating the porosity, extracting the interface topography of the coating, and analyzing structure characteristics such as internal defects; and finally, estimating the preparation process quality of the heat barrier coating by integrating the composition, thickness, porosity distribution, interface topography and the like. The method for three-dimensional representation of the stratified structure of the heat barrier coating provided by the invention is suitable for estimation and optimization of the spraying process quality of the heat barrier coating, so that the quality of the heat barrier coating can be well controlled.

The invention relates to a compressed sensing measurement matrix optimization method and system based on automatic encoder network. The method comprises the steps of obtaining original images as training data and dividing the training data into a plurality of image blocks through segment cutting processing; conducting sampling on the image blocks based on the preset sampling rate and the automaticencoder network, and generating initial reconstruction images; calculating the residual error value between the initial reconstruction images and the original images based on the deep residual errornetwork; combining the residual error value with the initial reconstruction image and generating a reconstruction result, and establishing a loss function based on the reconstruction images and the image blocks, and conducting training on the parameter in the automatic encoder network through the loss function, and using the automatic encoder network parameter after training completion as the compressed sensing measurement matrix. The invention is advantageous in that through the transformation of data dimensions via the automatic encoder, the process from collecting to reconstructing of images can be simulated and realized, wherein the parameter in the collecting process is the measurement matrix, which has good reconstruction quality.

The present invention discloses a distance compensation multiple-input-multiple-output array millimeter wave three-dimensional imaging device and method. The device comprises a single-channel emitterwhich generates linear frequency modulation millimeter wave signals and transmits the linear frequency modulation millimeter wave signals to a two-dimensional antenna array and a multi-channel receiver; the two-dimensional antenna array comprises an emission antenna and a receiving antenna which are distributed in a two-dimensional plane mode, the emission antenna radiates millimeter wave signalsto a free space, and the receive antenna receives scattering echo signals of a target object; the multi-channel receiver comprises a plurality of same receiver channels, the receiver of each channel performs orthorhombic down-conversion processing of the linear frequency modulation millimeter wave signals from the single-channel emitter and the echo signals from the receive antenna to obtain amplitude signals and phase signals of the echo signals and transmit the amplitude signals and the phase signals to a data processing system; and the data processing system employs the amplitude signals and phase signals of the echo signals to achieve the image reconstruction of the target object through adoption of the distance compensation.

The invention discloses a three-dimensional electrical capacitance tomography (ECT) imaging image reconstruction method which employs the measuring capacitance value obtained by a three-dimensional ECT imaging sensor as the projection data for image reconstruction. In an iteration reconstruction process, a threshold filtering mode is employed to perform artifact inhibition on an obtained reconstruction image; a filtering threshold regularly performs adaptive adjustment, and is selected to minimize the fuzzy measurement of a current image. The method effectively reduces reconstruction image artifacts, improves complexly distributed three-dimensional imaging effects, and realizes real three-dimensional ECT imaging image reconstruction.

The invention discloses an X ray multi-energy-spectrum CT (Computed Tomography) finite angle scanning and image iterative reconstruction method, which comprises the following steps that: applying theX rays of a plurality of energy spectrums to scan an object in a respective finite angel range, and obtaining the projection data of a tested object under each energy spectrum; on the basis of the projection data, independently reconstructing the image of each tested object under each energy spectrum; utilizing the consistency of the tested object and the complementarity of the reconstructed imageunder each energy spectrum, designing a combined regularization method to effectively remove an artifact caused by finite angle scanning in the reconstructed image under each energy spectrum, and calculating to obtain a high-quality reconstructed image; and therefore, applying an image domain decomposition method or a projection domain decomposition method to calculate to obtain the base materialimage or the dual-effect image of the tested object. The method is suitable for the image reconstruction of the multi-energy-spectrum CT under a finite angle scanning condition, the dosage of the X ray is lowered, scanning time is shortened, and the method has the advantages of high reconstructed image quality, high rate of convergence and the like.

The invention relates to a reconstruction algorithm for back projection weight cone-beam CT, belonging to the field of X-ray computed tomography. The reconstruction algorithm comprises the following steps of: introducing a ray back projection weight based on a filtering back projection reconstruction frame, acting the back projection weight based on distance and continuity of conjugate rays on filtered projection data during back projection so as to reconstruct a CT image, and performing certain compensation on Radon missing data to improve the intrinsic cone-beam artifact problem of an FDK (Feldkamp-Davis-Kress) algorithm under circular track and increase the reconstruction quality. According to the reconstruction algorithm disclosed by the invention, the frame based on the filtering back projection has simple and efficient reconstruction process and is insensitive to noise.

The invention discloses a CT (computed tomography) machine. The CT machine comprises a bulb tube, a ray box, a collimator and at least one pair of reference detectors, wherein the collimator is positioned in the ray box; the reference detectors are arranged in the ray box; the reference detectors and the bulb tube are respectively positioned on two sides of the collimator along a ray exiting direction; and the reference detectors are positioned in a penumbra field of rays of the bulb tube. The invention also provides a real-time monitoring method for the focal point of the bulb tube of the CT machine. The reference detectors are arranged in the ray box of the CT machine, so that the size and the position of the focal point of the bulb tube can be simultaneously monitored. Compared with a method for only monitoring the position of the focal point, the real-time monitoring method has the advantages that the accurate data of the position and the size of the focal point of the bulb tube can be acquired, and the quality of reconstructed images is improved.

The invention belongs to the technical field of X ray computerized tomography (CT), and particularly relates to a wide visual field cone-beam X ray oblique scanning three-dimension digital imaging method based on an algebraic reconstruction algorithm. The method comprises the following steps: acquiring two-dimension projection data of a scanned component in detector biasing wide visual field cone-beam X ray oblique scanning way; and rebuilding an image by the algebraic reconstruction algorithm to obtain a three-dimension computerized tomography image in a scanning area. Compared with the widevisual field cone-beam X ray oblique scanning three-dimension digital imaging method based on a filter back projection algorithm, the method can effectively inhibit structure aliasing between layers and obviously improve the quality of the rebuilt picture in the condition that the system hardware, the scanning view field and the scanning speed are not changed.

The invention discloses a multi-frame low-resolution image super-resolution reconstruction method based on a convex combination mode, belongs to the field of computer vision, mainly solves a problem of image processing precision and provides an optimization algorithm. The method comprises the following steps of applying joint motion estimation and super-resolution reconstruction methods to an input multi-frame low-resolution (LR) image sequence to obtain a joint high-resolution (HR) image Ijoint; applying a reconstruction method based on improved kernel regression to the input multi-frame LR image sequence to obtain a kernel regression HR image Ikernel; and applying a convex combination frame provided by the invention and deciding a fused region according to an airspace fusion rule, deciding frequency components that need to be fused according to a frequency domain rule and deciding weight distribution of fusion of two methods according to the global weight, and then fusing the HR images obtained by the two steps through the three aspects to obtain a final reconstructed HR image.

The invention discloses a super-resolution reconstruction method of multi-frame images. A steering kernel regression method is introduced separately to a clustering phase of an adaptive sparse learning model and a regularized reconstruction process; the two methods are enabled to complement each other's advantages and disadvantages, forming a reconstruction algorithm of better performance. Motionregistration is not required, and therefore, reconstruction errors due to registration errors are avoided and the quality of a reconstructed image can be significantly improved; the method is applicable to places including any motion mode. Compared with existing learning-based multi-frame super-resolution reconstruction methods, the simple efficient integrated reconstruction model that integratessteering kernel regression and sparse learning is provided which can make use of both global structural self-similar priori constraints and sparse constraints to carry out regressive estimation, so that edge details of an image can be better maintained, and image high-frequency distortion can be relieved; the method is not limited to motion amplitude or application scenes and accordingly is adaptive to complex environments.

An embodiment of the invention provides a scanning method for PET equipment, which is applied to PET console computer device. The scanning method comprises the following steps: receiving a first reconstructed image transmitted by reconstruction computer equipment, wherein the first reconstructed image is an image obtained by performing image reconstruction based on original scanning data; determining a first scanning time interval when determining that the first reconstructed image is abnormal, wherein the first scanning time interval is a time interval determined based on the original scanning data and counting rate information after abnormal data are deleted, and the original scanning data comprise scanning time information; generating a control command according to the first scanning time interval; and transmitting the control command to the reconstruction computer equipment, wherein the reconstruction computer equipment is used for selecting data corresponding to the first scanning time interval from the original scanning data according to the control command to reconstruct an image and obtain a second reconstructed image. The embodiment of the invention can effectively improve the quality of the reconstructed image; the method is simple, flexible and effective.

The invention discloses an image coding method, an image coding device, an image decoding method and an image decoding device based on a super-resolution technology. The image coding device is positioned at a sending end of an image, and comprises a downsampling module used for performing downsampling on the original image; a first compression module used for compressing the downsampled image forthe first time; a neural network used for performing super-resolution reconstruction on the image that is compressed for the first time to acquire the image that is reconstructed for the first time; afirst addition module used for subtracting the image that is reconstructed for the first time from the original image to acquire a first residual image; a second compression module used for compressing the first residual image for the second time; and a sending module used for jointly sending the compressed first residual image and the image that is compressed for the first time to a receiving end of the image. According to the methods and devices provided by the invention, the code rate is greatly reduced, and meanwhile, the quality of the reconstructed image also can be well enhanced by theapplication of the deep learning technology.

The invention relates to a high-efficiency image collection and compression method. The method includes: adopting an image detector array to acquire a quantized digital image; utilizing a lower sampling module to acquire thumbnail data of an original image; acquiring data of the original data after dimension reduction through a compression sensing module; sequentially outputting the thumbnail data and compression sensing dimension-reduction data through a compressed data output module according to a preset compression ratio to form code stream used for transmission or storage; after the compressed data are stored and transmitted, completing reconstruction of the original image through an image reconstruction module. By the high-efficiency image collection and compression method, complexity of conventional image compression is lowered effectively, and the defect that a reconstruction step is needed to realize image looking-back on the basis of conventional compression sensing data is overcome.

The embodiment of the invention discloses a depth map prediction method, a pixel detection method and related devices. The depth map prediction method comprises the following steps: the maximum value and the minimum value in the characteristic values of a current block are acquired; if the current block is an image block in a depth map, the characteristic values are original pixel values; or if the current block is a residual block obtained by predicating the image block in the depth map, the characteristic values are residual values; and if the absolute value of the difference between the maximum value and the minimum value of the characteristic values of the current block is greater than or equal to a first threshold, segment-wise DC coding is carried out on the current block according to the characteristic values of part of pixels of the current block, wherein the part of pixels of the current block do not include a first class of pixels in the current block, and the first class of pixels in the current block are pixels of which the characteristic value is greater than or equal to a second threshold in the current block. The scheme of the embodiment of the invention can improve the coding efficiency and the quality of vision-synthesized reconstructed images.

The invention discloses a motion estimation and adaptive video reconstruction method based on interpolation. In the time super-resolution video reconstruction process based on compression perception,for a sample learning reconstruction algorithm, the close correlation between the video image reconstruction quality and training samples is considered, and adaptive classification reconstruction is carried out according to the object exercise amount. According to the method, firstly, a coded image is acquired through a pixel continuous equal-length exposure mode, secondly, the motion area is segmented through utilizing a method based on coded image scattering characteristics, initial video images are rapidly acquired through utilizing an interpolation method, the exercise amount of each areais estimated based on the initial images, and lastly, corresponding training parameters are selected according to different motion amount to carry out classification reconstruction. The method is advantaged in that the scene motion information is rapidly acquired before image reconstruction, and the video image reconstruction quality is further improved.

The invention provides an ultrahigh-speed compression photographing device based on compression perception and streak camera principles. The device comprises a lens, a beam splitter, a spatial light modulator, a flat mirror, a streak camera with a built-in CCD, and a decoder. Firstly, captured image information is coded through the spatial light modulator. Then the streak camera is utilized for superposing image information at different time points and performing compression sampling. Finally reconstruction is performed on the image based on a compression perception principle. According to the device, two-dimensional imaging can be performed on an ultrafast process on the condition of definite imaging speed of the streak camera; the image information is output to a decoder; a two-dimensional image dynamic process (x-y-t) is reconstructed through a TwIST algorithm; and the imaging speed can reach 10-12 frame / second. The ultrahigh-speed compression photographing device utilizes single-time photographing measuring technology which can measure un-repeated ultrafast events.

The invention provides a coincidence processing method of a scintillation pulse event. The method comprises the following steps: S1, acquiring experimental data of the PET system; wherein the experimental data comprises time information that a pair of gamma photons with the same energy and opposite directions generated after positron annihilation arrives at two corresponding detectors and energy information that the gamma photons deposit in the detectors; wherein the time information is Ti1 and Ti2 respectively; wherein the energy information is Ei1 and Ei2; s2, a step S2 is performed; settingan energy window and a time window and acquiring a coincidence event; energy window setting E, time window setting to T, wherein the gamma photons meeting the following conditions are marked as a pair of true coincidence events: |Ti1-Ti2|<=T, Ei1<=E, and Ei2<=E. The method comprises the following steps: S3, carrying out weighting processing on the true coincidence events, and carrying out enhancement processing by adopting two formulas, S4, carrying out image reconstruction to obtain an image result, and S5, analyzing the image result. The weights of the coincidence events are distributed byutilizing the energy information, the quality of the reconstructed image is improved, and the adaptability is high.

The invention discloses a support set and signal value detection based video compressive sensing reconstruction method which mainly solves the problem of poor reconstructed image quality in the prior art. The method includes the implementation steps: (1) dividing a video sequence into reference frames and non-reference frames according to image groups; (2) dividing the reference frames and non-reference frames into non-overlapping macro blocks identical in size; (3) subjecting all the macro blocks to compressive sensing measurement; (4) utilizing measurement values as input and updating iteration variables of a reconstructed image; (6) updating a support set and a signal detection value according to updated iteration variables of the reconstructed image; (7) computing a residual error of the reconstructed image according to the signal detection value; (8) judging whether iteration is terminated or not according to constraint conditions of the residual error of the reconstructed image; (9) outputting a reconstructed image signal. The support set and signal value detection based video compressive sensing reconstruction method can improve reconstructed image quality effectively and can be utilized for video image processing.

The invention discloses a CT image reconstruction method and system based on a convolutional neural network, and the method comprises the steps: inputting projection data, carrying out the filtering processing and back projection processing of the projection data through a front-end convolutional neural network, generating tomographic image information, and processing the generated tomographic image information through a back-end convolutional neural network. And calculating the value of the whole network loss function by using the processed tomographic image and the real image, reversely feeding back the gradient information of the loss function layer by layer, and updating the parameter value of each network layer. According to the invention, the projection data is processed by using the multi-layer convolutional neural network, useful information in the projection data is fully mined, and the feature extraction and information expression capabilities of the overall neural network model are effectively improved; a full connection layer is prevented from being used in the neural network model, and the required neural network model parameters are few and easy to implement.

The invention discloses a multimedia image compression method based on a non-correlation chaos observation matrix. The method comprises the steps that a video to be detected is read; under the condition that compression ratio is determined, observation frequency is determined; a Logistic mapping system is selected to produce a chaos factor; through the chaos factor, the first row of the observation matrix is determined, and a sequential cycle method is used to acquire vectors of other rows of the observation matrix; a non-correlation factor is introduced to strengthen the column non-correlation of the observation matrix; for each element moving from the tail end to the front end, the element is multiplied by the non-correlation factor; each time the element moves, the element is multiplied by the non-correlation factor in a superposed manner; and QR decomposition is used to process the observation matrix to acquire an orthogonal observation matrix based on non-correlation chaos. Compared with similar algorithms, the method provided by the invention has the advantages that a storage space restriction is made up; sufficient column non-correlation is met; a multimedia image is well compressed; the quality of a reconstructed image is great; the method is robust for different multimedia images; and the method is highly practical.

The invention relates to a CT imaging system and an imaging method, belongs to the technical field of radiation imaging, and solves the problem that an existing CT system adopting static CT equipment needs a large number of radiation sources and detectors. The CT imaging system comprises a static imaging subsystem, a dynamic imaging subsystem and a rack, the static imaging subsystem and the dynamic imaging subsystem are both fixed to the rack, and the static imaging subsystem comprises multiple sets of static imaging units which are fixedly arranged. By fusing the static imaging subsystem and the dynamic imaging subsystem, the arrangement number of ray sources and detectors in static CT can be reduced, relatively complete projection data can be obtained in a data fusion mode, the quality of reconstructed images is improved, and relatively complete and ideal overall reconstruction results are obtained.

The invention discloses a multi-image encryption and decryption method based on Walsh transform and ghost imaging calculation, and the method comprises the steps: encryption: carrying out Walsh-Hadamard compression transform on a plurality of plaintext images, selecting some effective pixels to carry out Arnold scrambling transform, and then loading the light intensity of an object plane to a calculation ghost imaging light path in sequence, and encrypting a target image into a ciphertext sequence; and 2, decryption: in the decryption process, sucessfully decrypting a corresponding plaintext image from a correct key set through measurement key recovery, compression algorithm reconstruction, inverse Arnold transformation and inverse Walsh-Hadamard transformation. The image encryption method provided by the invention has the advantages of simple encryption operation, high decryption efficiency and high security.

The invention provides a linear scanning CL reconstruction method based on projection visual angle weighting. The method comprises the following specific steps: 1) data collection: carrying out the CL linear scanning of a to-be-measured object, and obtaining projection data; 2) performing image primary reconstruction: performing image reconstruction on the projection data collected in the step 1) by using a filtered back projection algorithm to obtain reconstructed images under different visual angles; 3) determining a back projection weighting coefficient: determining a reference value of the layered images according to the reconstructed images at different view angles in the step 2), calculating dissimilarity between the layered images at different projection view angles according to the reference value, and determining the back projection weighting coefficient according to the dissimilarity and adjustment parameters alpha and beta; and 4) performing back projection reconstruction: performing back projection reconstruction by using the back projection weighting coefficient obtained in the step 3), thereby solving the problems of incomplete projection data, image limited angle artifacts, layered image blurring and the like caused by the fact that the plate-shaped part is not suitable for full-angle scanning, reducing layered image aliasing, and improving the quality of the reconstructed image.

The invention discloses a compressed sensing and cleaning algorithm combined image reconstruction method and system, and the method comprises the steps: 1), searching an appropriate sparse domain fora sparse brightness temperature distribution image according to a compressed sensing principle, building an objective function, and solving a compressed sensing reconstruction image through an iterative algorithm; 2) performing Fourier transform on the compressed sensing reconstructed image, obtaining a frequency domain visibility function value of a sampling baseline position, subtracting the original image sampling value from the frequency domain value, and performing inverse Fourier transform to obtain a residual inversion image; 3) performing cleaning operation on the residual inversion image by adopting a cleaning method to obtain a residual reconstructed image; and 4) adding the compressed sensing reconstructed image and the remaining reconstructed image to obtain a final reconstructed image. The method provided by the invention aims at the problem of serious image aliasing under the condition of sparse sampling in passive microwave interference imaging, and improves the qualityof a reconstructed image.

The invention discloses a CT image sparse reconstruction method of an iterative evolution model. The method comprises the following steps: acquiring sparse projection data with a fixed projection angle range or a projection angle interval reaching a set value; Generating four initial solutions, namely a pseudo-inverse solution, a random value solution, a total 0 value solution and a total 1 valuesolution; obtaiing Four groups of solution sets through random walk; Performing validity constraint on each solution in each solution set; Evaluating each solution in each group of random solution setaccording to a fitness evaluation function; transferring each solution according to a novel iteration evolution model; Updating each solution according to a convex optimization-based iterative algorithm; Selecting the current optimal solution update of each group according to the fitness evaluation; And if so, selecting an optimal solution as a reconstruction result according to the fitness evaluation. The method can improve the quality of the reconstructed image, and reduces the reconstruction artifacts.

The invention discloses a matching tracking reconstruction method based on structured observation and direction guidance, which mainly solves the problem of inaccurate and low robustness of compressed sensing image reconstruction under low sampling rate, and the realization process is as follows: according to the sketch of the image , defines sketchable blocks and non-sketchable blocks; clustering based on sketch direction guidance is used for sketchable blocks, and grayscale clustering is used for non-sketchable blocks; when observing, sketch-based representation is used for smooth image block classes in non-sketchable blocks The structured observation of non-smooth image block class adopts multi-measurement vector observation; when reconstructing, the OMP algorithm based on sketch representation and direction guidance is used to obtain the reconstructed image. Compared with the OMP method under traditional observation, the present invention has important advantages With the advantages of high image quality and good robustness, it can be used for the reconstruction of natural images.

The embodiment of the invention provides a multi-angle free view angle data generation method and device, a medium and equipment, and the method comprises the steps: obtaining a plurality of synchronous images, and enabling the viewpoints of the plurality of images to be different; determining a virtual path, wherein the virtual path is a set of viewpoints for watching the to-be-watched area; pre-processing the plurality of images according to a virtual path and the viewpoint of each image in the plurality of images to obtain a pre-processed image set consistent with the virtual path; and generating multi-angle free view angle data based on the preprocessed image set. According to the technical scheme, the image quality can be guaranteed when the viewing angle is adjusted for viewing.

The invention provides a glasses removing method and device and communication equipment. The method comprises the following steps: segmenting an initial face image into a first image block, a second image block and a third image block; performing up-sampling operation and down-sampling operation on the second image block in sequence through the glasses removal shallow convolutional neural network, and reconstructing the image through convolution and pooling to obtain a reconstructed glasses-free image block consistent with the second image block in size; splicing the reconstructed glasses-free image block, the first image block and the third image block to obtain a complete glasses-free face image; according to the embodiment of the invention, the feature map with higher resolution is obtained through up-sampling, and the quality of the reconstructed image is improved; then feature extraction and reconstruction are carried out on the feature map through pooling, a double-branch convolution module and other operations, the process is a down-sampling process at the same time, and the size of the processed feature map is reduced to the size of an original map; the number of layers of the neural network is small, and therefore small calculation amount and high forward operation efficiency are kept.

The invention discloses a method and a device for processing a security check CT (Computed Tomography) reconstructed image. According to the method, projection data collected by a security check CT wide-row detector is subjected to inter-row expansion, the pixel height after expansion becomes half of the pixel height before expansion, and the expansion row data is obtained through correlation calculation according to two rows of projection data closest to the expansion row data; and iterative inter-row data expansion is performed on the projection data according to the row spacing and the actual pixel width until the pixel height of the expanded projection data is smaller than or equal to the actual pixel width. A CT image processing device comprises a data acquisition module, a wide-row CT projection data expansion module and a reconstruction module. The invention has the advantages that the expanded projection data is used for reconstruction without changing existing scanning conditions, the axial data sampling rate is increased, windmill artifacts are greatly reduced, and the quality of reconstructed images is obviously improved.