377results about How to "Improve reconstruction quality" patented technology

An intra prediction method and a device using the intra prediction method are provided. The intra prediction method includes the steps of: deriving a current prediction mode as a prediction mode of a current block; constructing neighboring samples of the current block with available reference samples; filtering the available reference samples; and generating predicted samples of the current block on the basis of the filtered available reference samples. The filtering step includes performing the filtering using the available reference sample located in the prediction direction of the current prediction mode and a predetermined number of available reference samples neighboring to the prediction direction of the current prediction mode.

The invention relates to a microscope in which a layer of the sample is illuminated by a plurality of thin strips of light (11) passed through a grid (34) and the sample is viewed (5) perpendicular to the plane of the strips of light. To record the image, the object (4) is displaced through the strips of light (11). At least three different images of the objects (4) are made at different phase angles. The images can be combined to form a single combined image.

A method and apparatus are disclosed that seek to improve the quality of service that is experienced during the transmission of a stream of packets across one or more paths. In particular, a transmitting node encodes a source stream of data (e.g., audio, video, etc.) into one or more sub-streams, and distributes those sub-streams onto multiple network transmission paths. In accordance with the illustrative embodiment of the present invention, the transmitting node evaluates the quality of service of a first network path that fails to provide a quality-of-service guarantee. When the quality of service of the first network path becomes unsatisfactory, the coding of one or more sub-streams that are being transmitted on a second network path is adjusted. In other words, the coding on a second channel is adjusted in response to the changing conditions on a first channel.

The invention relates to a three-dimensional reconstruction algorithm introducing an interactive operation for a user, belonging to the technical field of a computer multimedia. The method is as followings: the user utilizes a plurality of cameras to obtain a two-dimensional image Ii of a plurality of viewpoints of an object and extracts a contour map of the object; the computer obtains a visual shell model of a real scene according to a geometric parameter of every camera and the contour map, and turns the visual shell model into a point cloud form; a two-dimensional characteristic in the image is extracted to optimize a part of a real surface model corresponding to the characteristics; the user divides the image Ii into a region of a smoothness and the region of a subsidence; a nonuniform weighted graph is set up according to the areas selected above to find out a minimal cut of the image to obtain the optimized scene surface model. Combined with an auxiliary operation of the user, the three-dimensional reconstruction algorithm is capable of reconstructing the high-quality three-dimensional model of the object relatively fast according to the multi-view image which is shot of the actual object and the camera parameter corresponding to each image.

This invention relates to a method, system, transmitter, network element, receiver and software application for a system capable of point-to-multipoint transmission, wherein one or more data packets are transmitted from a transmitter to one or more receivers, wherein at least at one specific receiver of said receivers, a reception of repair data packets is required, wherein repair information is signaled to a repair server in order to trigger a transmission of said repair data packets, and wherein said repair information comprises information related to the number of transmitted data packets correctly received at said specific receiver. Said system may for instance be the 3GPP Multimedia Broadcast Multicast System, said transmission of data packets may for instance be controlled by the File Delivery Over Unidirectional Transport protocol and said signaling of said repair information may for instance be controlled by the Hypertext Transfer Protocol.

The present invention relates to a method for reconstructing a 3D image from 2D X-ray images acquired with an X-ray imaging system, said method comprising the steps of:a) receiving a set of 2D X-ray images of a region of a patient with said X-ray imaging system,b) computing an initial 3D image within the coordinate system of the X-ray imaging system by using at least part of said 2D X-ray images with their respective projective geometry data;c) projecting said initial 3D image on at least part of said 2D X-ray images and adjusting the respective projective geometry data of said images, said adjustment comprising registration of said images with the projection of the initial 3D image using an image-to-image registration technique;d) computing an updated 3D image using the complete set of 2D X-ray images with their respective adjusted projective geometry data.

Geometry of a tomosynthesis system including a detector and an x-ray source is determined using fiducial markers with non-determined positions. The geometry is determined by arbitrarily identifying at least two markers within an imaged volume, at different relative distances between the detector and the x-ray source, without having projections located on a straight line for all different source positions, and locating the projections of the markers within at least two images acquired of the imaged volume. The at least two images correspond to different positions of a focal spot of the x-ray source.

A single-photon or ultra-weak light multi-D imaging spectral system and method. In order to realize rough time resolution, a time-resolved single-photon counting 2D imaging system for forming color or grey imaging is provided. Moreover, in order to realize high-precision time resolution, the system comprises a light source, an imaging spectral measurement unit, an electric detection unit, a system control unit and an algorithm unit. The light carrying information of an object is imaged on a spatial light modulator and randomly modulated according to compressed sensing theory, emergent light of a grating is collected using a point or array single-photon detector, the number of photons and photon arrival time are recorded, and reconstruction is carried out using the compressed sensing algorithm and related algorithm of the spectral imaging. The system provides single-photon detection sensitivity, high time resolution and wide spectral range, and can be applied in numerous new high-tech industries.

The invention discloses a text image super-resolution reconstruction method based on a conditional generative adversarial network. According to the method, the structure of a common conditional generative adversarial network is modified, so that the method is more suitable for a super-resolution reconstruction task of a text image; Non-text binary segmentation image is used as extra training supervision information of a super-resolution reconstruction model, and text-text binary segmentation image is combined with extra training supervision information of a super-resolution reconstruction model; the non-text binary segmentation information constructs a loss function of the model to constrain the training of the model, so that the super-resolution reconstruction model is more concentrated in the text part in the image. Compared with a common image super-resolution method, the text image super-resolution reconstruction method disclosed by the invention fully and specifically utilizes theinformation of the text, and the quality of text image super-resolution reconstruction is effectively improved.

The invention proposes a method for compressed sensing video reconstruction based on a recursive convolution neural network, and the method mainly comprises the contents: a compressed sensing network (CSNet), a CSNet algorithm structure, a convolution neural network (CNN), a long-short term memory (LSTM) network, CSNet training, and compressed sensing video reconstruction. The process comprises the steps: extracting movement features through an RNN; extracting visual features through a CNN; carrying out the fusion of the movement features and the visual features; gathering all the extracted features through the LSTM network, and combing the features with a deduced movement in a hidden state to achieve the reconstruction. The method solves a problem that the video reconstruction quality is difficult to guarantee at a high compression rate through a conventional method, and an end-to-end training and non-iteration model is designed. The method improves the compression rate (CR) of a CR camera, improves the video reconstruction quality, reduces the bandwidth of data transmission, and is enabled to support a video application with a high frame rate.

The invention discloses a multi-energy-spectrum CT image reconstruction method based on projection estimation, wherein the method is used for reconstruction of density images of various base materials of a measured object. The method comprises the steps of directly reconstructing energy images of the measured object from collected multicolor projection data according to a traditional single-energy CT reconstruction method, calculating the line integral of each energy image in the radial directions of all other energy spectrums, estimating multicolor projections of a current energy spectrum in the directions, obtaining the multicolor projections with the consistent geometrical parameters, calibrating a various base material division function, dividing the estimated multicolor projections with the consistent geometrical parameters into the line integrals of the various base materials, and reconstructing the corresponding density images of the various base materials through the line integrals of the various base materials. The multi-energy-spectrum CT image reconstruction method based on projection estimation is simple, practical and suitable for multi-energy-spectrum CT image reconstruction when the geometrical parameters of the multicolor projections are not consistent. Compared with the prior art, the multi-energy-spectrum CT image reconstruction method has the advantage that high-quality images can be reconstructed only by estimating the multicolor projections with the consistent geometrical parameters through the measured projection data.

The invention provides a method and a device of multiple description video coding based on a relevance optimization rule, belonging to the field of video coding. The device comprises a coder, a decoder, a channel 1 and a channel 2, wherein the coder comprises a similarity optimization rule module, a preprocessor, an odd and even frame separator, a standard coder 1 and a standard coder 2; and the decoder comprises a standard decoder 1, a standard decoder 2, an odd and even frame interleaver and a postprocessor. The video coding method better keeps time domain relevance in a video subsequence by using the relevance optimization rule aiming at different motion information on the premise of guaranteeing the compression performance of each video sequence, thereby ensuring that a one-way decoder can more effectively restore lost information so as to improve the reconstruction quality of the one-way decoder.

An intra prediction method and a device using the intra prediction method are provided. The intra prediction method includes the steps of: deriving a current prediction mode as a prediction mode of a current block; constructing neighboring samples of the current block with available reference samples; filtering the available reference samples; and generating predicted samples of the current block on the basis of the filtered available reference samples. The filtering step includes performing the filtering using the available reference sample located in the prediction direction of the current prediction mode and a predetermined number of available reference samples neighboring to the prediction direction of the current prediction mode.

The invention relates to an image super-resolution reconstruction method, belongs to the image processing technology field and solves problems that the edge information of an image generated in the prior art is fuzzy, application to multiple magnification times cannot be realized and the reconstruction effect is poor. The method comprises steps that a convolutional neural network for training andlearning is constructed, and the convolutional neural network comprises an LR characteristic extraction layer, a nonlinear mapping layer and an HR reconstruction layer in order from top to bottom; inputted paired LR images and HR images are trained through utilizing the convolutional neural network, training of at least two magnification scales is performed simultaneously, and an optimal parameterset of the convolutional neural network and scale adjustment factors at the corresponding magnification scales are acquired; after the training is completed, the target LR images and the target magnification times are inputted to the convolutional neural network, and the target HR images are acquired. The method is advantaged in that the training speed of the convolutional neural network is fast,after training is completed, and the HR images at any magnification times in the training scale can be acquired in real time.

The invention relates to a residual-based ultra-resolution image reconstruction method, which specifically comprises the following steps of: first calculating residuals between original high-resolution images and images obtained by performing interpolation amplification on low-resolution images; then establishing sample pairs by using the characteristics of low-resolution image samples and corresponding image residuals, classifying the sample pairs by taking the low-resolution image samples as references and adopting K-averaging, and training each type of sample pair by adopting a K-singular value decomposition (K-SVD) method to obtain dictionary pairs of the low-resolution image samples and the image residuals; and finally selecting a dictionary pair according to a Euclidean distance between a test sample and a type center, calculating the weighted sum of image residuals reconstructed by each type with similar Euclidean distances with the test sample as a final reconstructed image residual, and obtaining a high-resolution image by combining interpolation results of the low-resolution images. Only the image residuals are required to be reconstructed, and the high-resolution image can be reconstructed by combining the interpolated images, so that edge detail reconstruction results of the high-resolution image are improved.

A method of temporal prediction using motion estimation is implemented in a sequence of digital images, in the context of coding in a bitstream comprising at least first and second layers linked to each other in a chosen hierarchical relationship of scalability. Provision is made for constructing at least one set of reference pixels for the temporal prediction on the basis of information from a prediction image of the second layer and complementary information from an image of the first layer corresponding temporally to the prediction image. For each block of the current group of the image of the current second layer, a search is made for at least one block of the reference set of pixels so constructed that is suitable for the temporal prediction; and at least one corresponding motion vector is determined.

A system for transmitting data packets representing a source signal across a packet data network is provided. The encoder comprises a first encoder (110) and a redundancy encoder (120). The redundancy encoding is generated with a bit rate continuously scalable, the bit rate being provided by a bit rate controller (142) that uses input from the network (130) and packet-loss rate information. At the decoder, recovery is performed by a parameter estimator based in part on information transmitted from the first encoder using information from previous and / or future blocks and in addition on redundant information. The method may be added to existing lossy source coding systems or may be used to enhance the quality of the reconstructed source signal even in scenarios without packet loss.

The invention discloses a real-time 3D reconstruction method based on a depth map. The method comprises steps that depth maps and an RGB color map of a shooting scene are obtained through utilizing adepth camera; each frame of the depth maps is processed as follows, the depth information is complemented, pixel points are then converted into a first type of three-dimensional coordinate points, anda normal vector of each pixel point is calculated; the first type of three-dimensional coordinate points corresponding to the depth map are converted into a second type of three-dimensional coordinate points; each of the second type of three-dimensional coordinate points corresponding to the depth map is assigned a voxel block, and a hash table is utilized to index the voxel blocks; an sdf valueof each voxel in the voxel blocks is updated through weighted fusion of the voxel blocks, and the scene surface is extracted; the texture information of the scene surface is obtained, and a surface normal vector of each voxel of the scene surface is calculated. The method is advantaged in that the reconstruction speed and reconstruction quality can be effectively improved, and the method is applicable to large-scale scene reconstruction.

Provided is a digital holography device having a simple configuration and an improved image quality, including: a light source that emits light, the light source being provided for supply of object light beams formed by radiation, transmission, scattering, reflection, or diffraction of the emitted light from a subject; an array device that splits the light emitted from the light source into two kinds of reference light beams having different phases in a plane perpendicular to a direction in which the light emitted from the light source travels; a CCD camera having an image-capturing plane on which two kinds of interference fringe patterns are recorded, the interference fringe patterns being formed by interferences between the two kinds of reference light beams, which have been produced by the array device, and the object light beams, which have been formed by radiation, transmission, scattering, reflection, or diffraction from the subject; and an image reconstruction device that generates a reconstructed image of the subject from the two kinds of interference fringe patterns recorded on the image-capturing plane.

The invention provides a video super-resolution processing method and device, and the method comprises the steps: obtaining each frame of image of a video, and inputting the image into a convolutionalneural network; Sequentially performing feature extraction, feature dimension reduction, nonlinear mapping and high-dimensional space mapping on the image through the convolutional neural network toobtain super-resolution features; Performing reconstruction according to the features obtained by feature extraction and the super-resolution features to obtain a super-resolution image; And finally,performing coding to form a super-resolution video code stream. The super-resolution processing of the video is realized through the convolutional neural network; Dimension reduction through features,nonlinear mapping and high-dimensional space mapping are carried out; the calculation complexity is reduced, the time complexity is reduced, the network learning difficulty is reduced and the complextexture of the output image is reserved by adopting the jump connection, so that the high reconstruction quality is realized while the real-time performance required by video processing is ensured, and the method has a very wide application prospect in the fields of video real-time transmission and compression, video restoration and the like.

A structured light pattern including a set of patterns in a sequence is generated by initializing a base pattern. The base pattern includes a sequence of colored stripes such that each subsequence of the colored stripes is unique for a particular size of the subsequence. The base pattern is shifted hierarchically, spatially and temporally a predetermined number of times to generate the set of patterns, wherein each pattern is different spatially and temporally. A unique location of each pixel in a set of images acquired of a scene is determined, while projecting the set of patterns onto the scene, wherein there is one image for each pattern.

A system and method for constructing fluoroscopic-based three dimensional volumetric data from two dimensional fluoroscopic images including a computing device configured to facilitate navigation of a medical device to a target area within a patient and a fluoroscopic imaging device configured to acquire a fluoroscopic video of the target area about a plurality of angles relative to the target area. The computing device is configured to determine a pose of the fluoroscopic imaging device for each frame of the fluoroscopic video and to construct fluoroscopic-based three dimensional volumetric data of the target area in which soft tissue objects are visible using a fast iterative three dimensional construction algorithm.

The invention relates to an image super resolution rebuilding method based on sparse representation and various residual. The image super resolution rebuilding method based on sparse representation and various residual includes the steps: first, calculating residual between images formed after the existing high resolution image and a low resolution image are amplified through an interpolation and calculating a high frequency portion and a low frequency portion of the residual; second, building a sample pair through low resolution image sample characteristic and corresponding the high frequency portion and low frequency portion of the image residual, utilizing a texture meta structure to classify the samples by regarding the low resolution sample as a standard and utilizing a singular value decomposition (KSVD) method to train each type of a sample to obtain a dictionary pair of the low resolution sample, the high frequency portion and low frequency portion of the image residual; finally, choosing the dictionary pair and combing the final image residual with the interpolation result of the low resolution image to obtain a high resolution image according to the texture meta structure type of the test samples. The image super resolution rebuilding method based on sparse representation and various residual just needs to rebuild the image residual, combines the interpolation image to rebuild the high resolution image and improves a rebuilding result of the high resolution image.

A method for reconstructing magnetic resonance parameters includes the steps: acquiring diffusion-weighted imaging K space data according to a random descending acquisition matrix; establishing a diffusion-weighted imaging mathematical model containing a dispersion tensor coefficient; defining an objective function, fitting the dispersion tensor coefficient by solving an optimization problem and adding sparse constraint; and solving the dispersion tensor coefficient according to the descending acquired data to minimize the objective function. In the method for reconstructing the magnetic resonance parameters, DWI (diffusion-weighted imaging) images do not need to be reconstructed, the dispersion tensor coefficient is directly obtained from the descending acquired K space data by means of calculation, mistaken fitting of the dispersion tensor coefficient caused by errors introduced in reconstruction of the DWI images is avoided, and error propagation is prevented. The invention further provides a system for reconstructing the magnetic resonance parameters.

The invention discloses a method, a device, and an apparatus for three-dimensional face reconstruction, and a compute readable storage medium. The method comprises: marking feature points of a face reference image in a face reference image set, performing global contour deformation on a reference model based on pixel coordinates of the feature points, and performing photometric normal reconstruction on a deformed primary face model by a photometric stereo technology based on a normal, based on a surface normal vector of the reconstructed target face model, using a mesh deforming technology toiterate to generate a target face model by the optimized normal. The invention provides a robust three-dimensional face reconstruction technology based on the photometric normal, with help of unrestrained image sets and reference models, and combined with the photometric stereo technology and a mesh deformation technology, restriction on face reference images is reduced, and advantages of the twotechnologies are used. The method prevents defects of using one technology alone to reconstruct faces, improves quality of three-dimensional face reconstruction, and realizes robust three-dimensionalface reconstruction with high accuracy.

The invention relates to a high spectrum, true color image and point cloud complementary indoor reconstruction method and system, belongs to methods including scene adaption, SVM based wave spectrum identification, high spectrum and true color image based cloud point closed-loop correction and distance calculation and greedy triangles, and solves the problem that a cloud point obtaining device isnot sensitive to material, illumination, heat and the like and the problem in full automation of 3D reconstruction. The method comprises the steps that 1) initial data is obtained; 2) a deletion position of the cloud point is identified from high spectrum data; 3) cloud information of 3D points is corrected preliminarily; 4) a repair result is corrected according to a true color image; and 5) thepoint cloud is filtered, splicing is automatic, the point cloud is made into a patch, and texture is mapped. The high spectrum and true color image information is used to repair the point cloud, disadvantages of hardware are overcome, the reconstruction quality is improved, 3D indoor reconstruction is more stable and more practical via realization of adaption and full automation, and the method and system have wider application prospects.

The invention discloses a super-resolution reconstruction method based on a multi-path deep convolutional neural network. The super-resolution reconstruction method comprises the following steps: acquiring a total training set; performing image preprocessing on the total training set; preparing a test set; reconstructing an image by using a convolutional layer of the convolutional neural network.According to the multi-path convolutional neural network structure provided by the invention, multiple branches are added to an original single-path neural network, so that image features with different scales can be processed by convolution kernels with different amounts, and reconstruction quality and visual effects are both improved compared with those of an original method while total parameter quantities are not increased.

The invention relates to a cone beam CT (Captive Test) iterative reconstruction algorithm and especially relates to a creating method of cone beam CT iterative reconstruction algorithm projection matrix. The invention provides a projection matrix characterization method based on a finite element model and a Radon operator for the high-precision characterization problem of the cone beam CT iterative reconstruction algorithm projection matrix. The invention provides a new projection matrix characterization method by combining the respective characteristics of a ray coverage model and a basis function model and fully considering the mathematic characterization of various physical processes of the projection according to the ray projection rule beginning from the mathematic characterization of a continuous three-dimensional natural image, so that the projection process is fully characterized. As shown in a test result, the method disclosed by the invention effectively improves the characterization precision and reconstruction quality of the model.

The invention discloses a compressive sampling-based (CS-based) video coding / decoding system and a method thereof and relates to the technical field of image compression. The video coding / decoding system consists of a CS-based video coder subsystem (100) and a CS-based video decoder subsystem (200), wherein the CS-based video coder subsystem (100) is used for realizing compressive coding on an input video sequence image at a video sending end to generate a video compression code stream, and the CS-based video decoder subsystem (200) is used for realizing decoding on the received video compression code stream at a video receiving end to reconstruct an original video sequence image. According to the CS-based video coding / decoding system disclosed by the invention, the compression efficiency and the quality of the reconstructed image can be effectively increased while low-complexity coding is performed, and the application requirements of a wireless video camera, a wireless video monitor and the like are met.