54results about How to "Guaranteed reconstruction quality" patented technology

The invention provides a 3G video transmission method and system based on compression sensing in a video monitoring scene, and belongs to the field of data compression and video transmission. After the format of video information collected by a camera is converted into the YUV format through color space conversion, H.264 coding is conducted, compression processing of compression sensing is conducted on coded data, video data streams obtained through compression are transmitted in a 3G network according to the RTSP protocol and the RTP / RTCP protocol, and reconstruction processing of compression sensing and the H.264 decoding process are conducted at a terminal so that restored video information can be obtained. Compression sensing is combined with the RTSP protocol and the RTP / RTCP protocol to be used in the 3G video transmission system, and therefore the system can meet the requirement for instantaneity of video transmission from the aspects of control and data, the system has the self-adaption video coding speed and network transmission speed regulation function, the size of the data transmitted through a network can be greatly reduced, broadband resources are saved, and the transmission efficiency of the system is improved.

The invention discloses a frequency domain registration and convex set projection-based super-resolution reconstruction method. The method comprises the following steps of: reading a multi-frame image, selecting a reference frame serving as registration, and registering the image to obtain relative displacement of each frame of image, wherein a frequency domain-based rapid registration algorithm is used as the registration algorithm; performing super-resolution reconstruction on the multi-frame image by using a convex set projection method according to the relative displacement of each frame; and finally outputting the image, wherein a video sequence sampling speed is high and the displacement of neighboring frames is small, so that by the method, the registration process is simplified, and the translational motion is estimated by using intermediate frequency domain information of the image and the rotation angle is not estimated to greatly reduce the amount of operation of the registration algorithm. Meanwhile, by the method, the convex set projection reconstruction method is simplified, the entire amount of operation of the super-resolution algorithm is greatly reduced by only using high-brightness amplitude constraint of the image and apriori constraint of a Gaussian point spread function, and the reconstruction quality can be guaranteed.

The invention relates to an image super-resolution method based on a channel attention mechanism and multilayer feature fusion, and the method comprises the steps of directly extracting the original features of a low-resolution image at the beginning of a residual branch by using a single-layer convolutional layer based on deep learning; using six cascaded convolutional circulation units based ona channel attention mechanism and multi-layer feature fusion to extract accurate depth features; carrying out upsampling on the depth features through a deconvolution layer, and carrying out dimensionality reduction on the upsampled features through a single-layer convolution layer to obtain a residual error of the high-resolution image; carrying out up-sampling on the low-resolution image by using a bicubic interpolation method in a mapping branch to obtain mapping of the high-resolution image; and adding the mapping and the residual of the high-resolution image pixel by pixel to obtain a final high-resolution image. The method is reasonable in design, fully considers the difference between the feature channels, efficiently utilizes the hierarchical features, and maintains a higher operation speed while obtaining higher accuracy.

The invention discloses a method for distributing video data. At least two video file copies with different code rates are prestored. The method comprises the following steps: receiving a client video access request which carries a network bandwidth parameter or specified video code rate; selecting a video file copy with the proximal network bandwidth parameter or specified video code rate; and when the difference between the code rate of the selected video file copy and the specified video code rate is less than a preset value, transmitting data of the selected video file copy to the client. The invention also provides a corresponding system for distributing video data. According to the invention, under the conditions of different network bandwidths, the contents of the optimum data bulk are automatically selected and distributed according to different bandwidths of users, so that the video can be played smoothly with high quality. The invention only needs to store a plurality of limited video file copies, thereby saving storage space and cost. In addition, on the premise of satisfying requirements of the user, the invention can realize the function of self-adaption code rate distribution.

The invention discloses a CT (Computed Tomography) image reconstruction method based on a variational inequality at a sparse sampling angle. The CT image reconstruction method comprises the following steps: (1) carrying out equal-angle interval projection scanning within an angle range of 0-180 degrees by aiming at a fan-beam CT to obtain spare projection data y, wherein an angle interval in a projection direction is between 2 and 6 degrees; (2) calculating a projection matrix A through an X-ray source, a detector and the position information of an object to be reconstructed; (3) according to the projection data y obtained in the step (1) and the projection matrix A obtained in the step (2), simultaneously introducing the sparsity and the non-negativity of an image gradient to serve as priori knowledge to obtain a reconstruction model of an image reconstruction problem at the sparse sampling angle; (4) converting the reconstruction model in the step (3) into a variational inequality form; and (5) solving the variational inequality in the step (4) to obtain a reconstructed image. On the premise that image reconstruction quality is guaranteed, reconstruction convergence speed can be quickened, and single-iteration time is shortened.

The invention belongs to the technical field of compressed sensing and distributed type video coding, and provides a multi-viewpoint distributed type video encoding and frame arranging method based on compressed sensing, wherein the multi-viewpoint distributed type video encoding and frame arranging method based on compressed sensing is suitable for nonlinear line distribution, the quality of side information between viewpoints is improved, and then the quality of decoded video frames is further improved. According to the technical scheme, an encoder part and a decoder part are included in the multi-viewpoint distributed type video encoding and frame arranging method based on compressed sensing. The encoder is adopted in the encoding step of key viewpoints and non-key viewpoints, wherein firstly, according to the arrangement sequence of the viewpoints, one key viewpoint, one key viewpoint and one non-key viewpoint are sequentially arranged, and frames in each viewpoint are arranged in the mode that key frames and non-key frames are arranged at intervals; secondly, the frames are divided into two kinds, one kind of frames are the key frames of the key viewpoints and the non-key viewpoints, the key frames are encoded according to the compressed sensing theory, and the non-key frames are encoded through the distributed type video coding method. The multi-viewpoint distributed type video encoding and frame arranging method based on compressed sensing is mainly used in the compressed sensing and distributed type video coding.

The invention discloses a three-dimensional video encoding method based on depth image rendering. The encoding method disclosed by the invention comprises the following steps of: projecting an original left-viewpoint colorful image to a right viewpoint from a left viewpoint through a method based on the depth image rendering, so as to obtain a residual image of an rendered image of an original right-viewpoint colorful image and the original left-viewpoint colorful image; carrying out pre-processing operation on the residual image and encoding the original left-viewpoint colorful image and the pre-processed residual image; and finally, obtaining a finally-reconstructed image of the decoded right-viewpoint colorful image according to the rendered image of the decoded right-viewpoint colorful image and the decoded residual image. According to the invention, on the premise of ensuring the higher reconstruction quality of the right-viewpoint colorful image, the encoding efficiency of a three-dimensional video is greatly improved.

The invention provides a bistatic ISAR sparse aperture imaging method, which relates to the technical field of radar signal processing and mainly solves the problems of low imaging resolution and longoperation time and the like of the bistatic ISAR under a sparse aperture condition. According to the realization process, a bistatic ISAR echo model is built, the echo after translation compensationis further subjected to Doppler displacement compensation, and full-aperture echo data are obtained; the echo is subjected to sparse representation, and a compressed sensing-based bistatic ISAR sparseaperture echo model is built; the whole two-dimensional echo data are blocked, and each pixel of a target image is assumed to obey Gaussian priori to build a sparse Bayesian model; a fast edge likelihood function maximization method is used to solve and obtain a high-quality target image; the target images corresponding to the solved blocks of echo are combined to a whole two-dimensional image, and a reconstructed target image is obtained. Through the technical scheme of the invention, while the reconstruction quality is ensured, the operation efficiency can be improved at the same time.

The invention discloses a compressed sensing based color imaging device and a compressed sensing based color imaging method. The compressed sensing based color imaging device comprises a PC (personal computer), a DLP (digital light projector), a target image, a single-pixel photon detector and a data collection and control module. The PC, the DLP and the single-pixel photon detector are connected with the data collection and control module. The PC generates a two-dimensional random binary projected intensity image. The data collection and control module controls the DLP to perform binary intensity image projection on the target image under red, green, blue and white structured light sequentially to acquire digital signals of the red, green, blue and white structured light respectively and send the digital signals to the PC, reconstructs the signals of the red, green, blue and white structured light through compressed sensing to acquire four monochrome gray level images, and performing color fusion on the four monochrome gray level images to acquire a final color image. The compressed sensing based color imaging device and the compressed sensing based color imaging method have the advantages that light filters are not needed, and simple structure and low cost are achieved; sampling frequency can be reduced, and color fidelity can be guaranteed without reduction of image resolution.

The invention discloses an image reconstruction method and device based on a computer tomography imaging system. The method comprises steps of acquiring acquisition parameters in a scanning protocol, and determining reconstruction parameters used for image reconstruction according to the acquisition parameters; according to the acquisition parameters and / or reconstruction parameters, determining a target reconstruction algorithm and according to the target reconstruction algorithm, generating a reconstruction protocol; and based on the acquisition parameters, acquiring scanning data of a to-be-detected object, and according to the scanning data and the reconstruction protocol, carrying out image reconstruction. According to the invention, through the acquisition parameters and the reconstruction parameters, the target reconstruction algorithm is determined, so linkage between the reconstruction parameters and the acquisition parameters can be achieved; according to the target reconstruction method, a reconstruction protocol adaptable to the current scanning protocol is generated, so problems of algorithm idling and repeated calculation of parts of algorithms caused by use of fixed parameters and fixed topology structures in the current reconstruction protocol are overcome and efficiency of image reconstruction is effectively improved.

The invention discloses an alternate iterative algorithm for temperature field and concentration field reconstruction based on spectral absorption. First, a linearization scheme is provided for calculation of a concentration field, an original system is semi-linearized, and thus the nonlinearity of the original system is weakened; then, an alternate iteration scheme is provided for the temperature field and the concentration field to be measured, and a half of the number of unknown variables is reduced in each iteration; finally, a penalty term is introduced through an optimization technology to overcome the morbidity in the reconstruction process, regularization parameters can be adjusted automatically in each iteration, and it is guaranteed that an iterative solution converges to an exact solution. According to the alternate iterative scheme, on the premise of maintaining reconstruction precision, the computing time can be shortened remarkably, and the alternate iterative algorithm is suitable for large-scale field inversion computation.

The invention discloses a method for quickly selecting the depth of an H.265 / HEVC inter-frame encoding unit. The method comprises the following steps: firstly, applying a top end skipping idea to method optimization by exploring the texture complexity of the encoding unit and the incidence relation between encoder quantization parameters and the depth selection of the encoding unit, designing an initial depth division prediction strategy of the encoding unit based on the texture complexity and the quantization parameters, and skipping the division of unnecessary large-size encoding units; then, applying an advanced termination idea to the method optimization by using the relevance among the depth selection of the encoding unit, the encoding code rate and the encoding distortion, designing a depth division termination selection strategy of the encoding unit based on code rate and distortion, and solving the problem of deciding division or not through an offline trained classifier. By adoption of the technical scheme of the invention, while the division precision of the encoding unit is ensured, the encoding time of 34.56% can be saved on average.

The invention discloses an adaptive de-noising method for Fourier ptychographic microscopy, and the method comprises the steps: taking an image under the condition that all light sources are turned off, and enabling the image to serve as a background light image Ib; sequentially turning on all LED units and taking a group of illuminating light images (shown in the description), enabling all illuminating light images to be subtracted by the background light image Ib, and obtaining a group of no-background-light images (shown in the description); carrying out the de-noising operation of all dark field images of the no-background-light images (shown in the description) according to a threshold value Tk, and gradually increasing the threshold value Tk from zero till the number of non-zero pixels in the dark field images is less than the half of the total number of pixels of the images; finally obtaining a group of dark field no-noise images (shown in the description), and calculating the ratio G of the mean values of the images (shown in the description), generated through an initial solution, and the dark field no-noise images in an iterative process of the Fourier ptychographic microscopy; employing the dark field no-noise images for updating if G is greater than 0.5 and less than 2, or else, carrying out no updating of the spectrums corresponding to the dark field no-noise images. The method improves the quality of the image reconstructed through the Fourier ptychographic microscopy, and stably and precisely constructs a large-view-field high-resolution image.

The invention provides a remote sensing image super resolution method based on dictionary learning, comprising the following steps: S1, performing dictionary learning of corresponding feature types based on an image library of different feature types; S2, identifying the feature type of an original image; and S3, carrying out an image super resolution reconstruction process using a corresponding feature type dictionary. The invention puts forward a super resolution method based on dictionary learning using the idea of feature classification. As the number of atoms in a dictionary used in the reconstruction process is reduced greatly, the dictionary scale is reduced greatly. The proportion of effective atoms in the classification dictionary used is increased significantly, so that the number of effective atoms in the dictionary used in the reconstruction process is almost the same as a general dictionary, and the speed of reconstruction is improved significantly while the quality of reconstruction is ensured. The process only involves the adjustment of parameters, the computation efficiency is high, and the reconstruction quality of remote sensing images is high.

The invention discloses a CT and X-ray luminescence computed dual-mode synchronous tomography method. Small animal X-ray luminescence computed tomography and computed tomography (CT) can be achieved synchronously, and a more accurate imaging object anatomical structure and molecular imaging information are provided. It is guaranteed that an XLCT projection and a CT projection are collected at the same location and same time by reducing the CT projection angle number and setting the CT projection angle number to be the same as that of the XLCT or be in a diploid relation with that of the XLCT; meanwhile, by means of a compressive sensing principle, a few-angle reconstruction algorithm achieving image total variation minimization is adopted for the CT projection, and a precondition matrix and wavelet transformation are introduced into XLCT iteration reconstruction so as to reduce ill conditioning of reconstruction. According to the CT and X-ray luminescence computed dual-mode synchronous tomography method, synchronous tomography of CT / XLCT can be achieved, dual-mode tomography processes are simplified, the experimental difficulty is reduced, the imaging time is reduced, the experimental failure rate is decreased, a radiation dosage is decreased, an error caused by objecting moving in the imaging process is avoided, and image registration and fusion difficulties are reduced.

The invention discloses a human body geometric reconstruction method based on Euler field deformation constraint, and the method comprises the following steps: taking a target human body grid as a data source, extracting human body feature points through employing a random forest regression method, and obtaining a target human body skeleton node through calculation; driving the skeleton to deformby using the feature points, and driving the grid to deform by using an LBS method, so that the attitudes of the template model and the target model are roughly aligned; on the basis of a local fieldgenerated by the HRBF, performing rigid transformation driven by local shape difference; performing regional expansion and contraction based on local shape driving; based on SDF, optimizing a target energy function based on target data constraint, human body deformation constraint and level set characteristic constraint in an Euler field, calculating a deformation vector field to obtain an SDF deformation result, and according to the SDF deformation result, driving a template grid to be aligned with the shape of a target model, so that human body reconstruction is realized. According to the method, the robustness and efficiency of three-dimensional human body structural reconstruction are improved, and the requirement of performing human body reconstruction on rough data is met.

The invention discloses a self consistency based parallel magnetic resonance imaging quick reconstructing method. Based on an SPIRiT frame, aiming at a JTV and JL1 composite regular term contained parallel imaging reconstruction problem, the invention proposes a quick reconstructing method employing Descartes sampling. In the prior art, scanning time can be reduced distinctively by adopting the magnetic resonance imaging technology; however, the reconstructing time of a prior reconstructing algorithm is long. The method provided by the invention utilizing a split Bregman technology for dissolving a prior problem into a plurality of sub problems that are easy to solve. Experiment simulation results show that the new algorithm is greatly improved in convergence speed under the premise of ensuring reconstruction quality compared with a prior NLCG algorithm.

The invention discloses a rapid chemical exchange saturation transfer imaging method and system, and the method determines that a current image employs a first downsampling rate or a second downsampling rate through judging the relation between the mean signal intensity of the image and a preset signal intensity threshold value; The method solves a problem that a low signal to noise ratio image is difficult to achieve a better reconstruction result because all CEST images employ the same downsampling rate, achieves the adaptivity of a CEST imaging undersampling mode, improves the CEST imaging speed, and guarantees the reconstruction quality of an CEST image.

The invention discloses a distributed video compression sensing reconstruction method based on time correlation. On the basis of establishing a distributed video compression sensing model without feedback, the method makes full use of the time correlation between non-key frames of video frames, performs pattern discrimination on each block of the non-key frames, and transmits the non-key frames with the residual encoding of the previous frame; and during the reconstruction, the reconstruction mode with edgeless information is adaptively selected according to the result of the mode discrimination, and the block of each non-key frame is reconstructed, thereby avoiding unnecessary calculation of the side information. The method utilizes the previous frame of the current frame for reconstruction instead of simply using the key frame as a reference frame for reconstruction, so that the time correlation between the video frames can be effectively utilized. The method can guarantee the reconstruction quality of the video at the same total sampling rate, while reducing the system reconstruction time, and no feedback channel is used.

The invention discloses a sparseness estimation-based distributed video compressed sensing fast reconstruction method. According to the method, based on relevance between distributed video frames, sparseness estimation is carried out for a measured value of compressed sensing, the estimated sparseness is substituted in an orthogonal matching pursuit algorithm for reconstruct compressed video images, and a great deal of calculation is reduced. The method accelerates reconstruction of the compressed sensing images on the premise of guaranteeing that quality of the reconstructed images is mainly unchanged, and meets the requirement of real-time property of video information better.

The invention discloses a fast CT reconstruction method of multi-scale sparse projection data. The method comprises the specific steps of collecting sparse angle projection data; carrying out down-sampling calculation by using original projection data and a corresponding projection matrix; reconstructing the original projection data by using a convex optimization method; reconstructing obtained down-sampling projection data by using a convex optimization iteration method and a constraint of a regularization term; fusing the reconstruction result; and utilizing a reconstruction algorithm for reconstruction again by employing the fused image as an initial value. According to the method, the sparse angle projection data is collected, reconstruction is carried out on an existing regularization constraint-based convex optimization algorithm, and the speed of the reconstruction algorithm is increased through fusing two reconstruction results on the basis of ensuring the original algorithm reconstruction quality, so that fast CT reconstruction of the multi-scale sparse projection data is achieved. The fast CT reconstruction method of the multi-scale sparse projection data can be widely used for the technical field of CT image reconstruction.

The invention discloses a convolutional neural network image super-resolution reconstruction method fused with a bionic vision mechanism, and the method comprises the steps: firstly carrying out the saliency region detection of a remote sensing image through employing a saliency detection method for simulating a human vision attention mechanism; secondly, for the salient region, performing super-resolution reconstruction by adopting an image super-resolution reconstruction method based on a convolutional neural network; and finally, carrying out super-resolution reconstruction on the non-salient region by adopting a bicubic interpolation method. Compared with an existing image super-resolution reconstruction method based on a convolutional neural network, the method provided by the invention can quickly perform super-resolution reconstruction on the image, and is suitable for occasions with strict real-time requirements.

The invention discloses a 3D-HEVC rapid transcoding method based on an unbalanced quadtree. Correlation between coding unit division depth information in coded high resolution bit streams and CTU (Coding Tree Unit) division depth in dimension-reduced videos, and the correlation between the CUT maximum division depth of texture graphs and depth graphs are utilized, a CTU division depth selection range is reduced, and an unnecessary division depth rate-distortion cost computing process is skipped, so a 3D-HEVC transcoding process is accelerated. Through adoption of the technical scheme providedby the invention, the video quality is ensured, and 25.88% of coding time can be reduced averagely.

The invention provides a 3G video transmission method and system based on compression sensing in a video monitoring scene, and belongs to the field of data compression and video transmission. After the format of video information collected by a camera is converted into the YUV format through color space conversion, H.264 coding is conducted, compression processing of compression sensing is conducted on coded data, video data streams obtained through compression are transmitted in a 3G network according to the RTSP protocol and the RTP / RTCP protocol, and reconstruction processing of compression sensing and the H.264 decoding process are conducted at a terminal so that restored video information can be obtained. Compression sensing is combined with the RTSP protocol and the RTP / RTCP protocol to be used in the 3G video transmission system, and therefore the system can meet the requirement for instantaneity of video transmission from the aspects of control and data, the system has the self-adaption video coding speed and network transmission speed regulation function, the size of the data transmitted through a network can be greatly reduced, broadband resources are saved, and the transmission efficiency of the system is improved.

The invention discloses an under-sampling hologram compression holography multi-scale self-focusing reconstruction method and system, and belongs to the technical field of digital holography. The method comprises the steps of carrying out down-sampling operation on holograms generated by an object at different positions in an optical structure through the irradiation of a light source, and obtaining the down-sampling holograms at different positions; based on a CS algorithm, reconstructing the down-sampling holograms at the different positions to obtain amplitude images at the different positions; estimating a focal plane position according to the amplitude images at different positions by adopting an evaluation method of a characteristic value self-focusing algorithm; and based on a TwIST algorithm, reconstructing the down-sampling hologram at the position of the focal plane. According to the invention, an EIG-AF-CS self-focusing algorithm is provided for realizing self-focusing under an under-sampling condition, and is combined with twin-free image compression reconstruction, so that the reconstruction quality is ensured, and the problems of twin image interference and lack of self-focusing capability in traditional compression holographic reconstruction are effectively overcome.

The invention discloses a video encoding complexity adaptive regulation method and device. The method comprises the following steps: pre-classifying macro block encoding modes; calling a complexity control module, preliminarily calculating a calculation complexity required by encoding according to the pre-classified encoding modes, comparing the pre-estimated calculation complexity with a target calculation complexity, and entering into subsequent video encoding procedures for performing movement estimation and mode selection when the pre-estimated calculation complexity meets the given target calculation complexity; otherwise, regulating the macro-block-grade calculation complexity until the target complexity requirement is reached. The invention also discloses a video encoding complexity adaptive regulation device. The device comprises a macro block pre-classification unit, a complexity estimation unit and an encoding complexity control unit. According to the method and device provided by the embodiment of the invention, when a power supply capability of a power source is insufficient, the encoding parameters can be regulated actively, the encoding complexity can be reduced, and simultaneously the quality of reconstructed video can also be ensured.