171 results about "Super resolution image reconstruction" patented technology

A method and system are presented for real time Super-Resolution image reconstruction. According to this technique, data indicative of a video frame sequence compressed by motion compensated compression technique is processed, and representations of one or more video objects (VOs) appearing in one or more frames of said video frame sequence are obtained. At least one of these representations is utilized as a reference representation and motion vectors, associating said representations with said at least one reference representation, are obtained from said data indicative of the video frame sequence. The representations and the motion vectors are processed, and pixel displacement maps are generated, each associating at least some pixels of one of the representations with locations on said at least one reference representation. The reference representation is re-sampled according to the sub-pixel accuracy of the displacement maps, and a re-sampled reference representation is obtained. Pixels of said representations are registered against the re-sampled reference representation according to the displacement maps, thereby providing super-resolved image of the reference representation of said one or more VOs.

The invention discloses a super-resolution reconstruction method based on a conditional generative adversarial network, and the method specifically comprises the steps: making a low-resolution image and a corresponding high-resolution image training set by using a disclosed super-resolution image data set; constructing a conditional generative adversarial network model, using dense residual blocksin the generator network, and realizing super-resolution image reconstruction at the tail end of the generation network model by using a sub-pixel up-sampling method; inputting the training image setinto a conditional generative adversarial network for model training, and enabling a training model to converge through a perception loss function; carrying out down-sampling processing on the imagetest set to obtain a low-resolution test image; and inputting the low-resolution test image into the conditional adversarial network model to obtain a high-quality high-resolution image. The method can well solve the problems that a super-resolution image generated by a traditional generative adversarial network looks like clear, and evaluation indexes are extremely low, and meanwhile, the problems of gradient disappearance and high-frequency information loss are relieved through a dense residual network.

The invention provides a feature extraction and matching method and device for a digital image based on PCA (principal component analysis), belonging to the technical field of image analysis. The method comprises the following steps of: 1) detecting scale space extreme points; 2) locating the extreme points; 3) distributing directions of the extreme points; 4) reducing dimension of PCA and generating image feature descriptors; and 5) judging similarity measurement and feature matching. The device mainly comprises a numerical value preprocessing module, a feature point extraction module and a feature point matching module. Compared with the existing SIFI (Scale Invariant Feature Transform) feature extraction and matching algorithm, the feature extraction and matching method has higher accuracy and matching speed. The method and device provided by the invention can be directly applied to such machine vision fields as digital image retrieval based on contents, digital video retrieval based on contents, digital image fusion and super-resolution image reconstruction.

The invention relates to a circuit-board element installing / welding quality detection method based on super-resolution image reconstruction. The method comprises the steps of utilizing a camera array and the motion of a conveyor belt to perform super-resolution image reconstruction on an area to be detected of a circuit board, and judging whether the installation and welding of elements are qualified according to reconstructed high-resolution images of the area to be detected on the circuit board. A detection system for realizing the method consists of a central server and a plurality of super-resolution detection ends, wherein the central server is connected with every super-resolution detection end; the super-resolution detection ends acquire the images of the area to be detected of the circuit board at detection points, perform super-resolution image reconstruction and then transmit the images to the central server; and the central server matches every high-resolution image in the area to be detected of the circuit board, which is acquired at the detection points, with a corresponding standard template, so as to detect unqualified elements or welding spots.

The invention discloses a multi-task super-resolution image reconstruction method based on KSVD dictionary learning, which mainly solves the problem of relatively serious quality reduction of the reconstructed image under high amplification factors in the existing method. The method mainly comprises the following steps: firstly, inputting a training image, and filtering the training image to extract features; extracting image blocks to construct a matrix M, and dividing the matrix M into K classes to acquire K pairs of initial dictionaries H1, H2...Hk and L1, L2...Lk; then, training the K pairs of initial dictionaries H1, H2...Hk and L1, L2...Lk into K pairs of new dictionaries Dh1, Dh2...Dhk and Dl1, Dl2...Dlk by utilizing a KSVD method; and finally, carrying out super-resolution reconstruction on the input low-resolution image by utilizing a multi-task algorithm and the dictionaries Dh1, Dh2...Dhk and Dl1, Dl2...Dlk to acquire a final reconstructed image. The invention can reconstruct various natural images containing non-texture images such as animals, plants, people and the like and images with stronger texture features such as buildings and the like, and can effectively improve the quality of the reconstructed image under high amplification factors.

This invention presents a super-resolution image reconstruction from a sequence of aliased imagery. The sub-pixel shifts (displacement) among the images are unknown due to uncontrolled natural jitter of the imager. A correlation method is utilized to estimate sub-pixel shifts between each low resolution aliased image with respect to a reference image. An error-energy reduction algorithm is derived to reconstruct the high-resolution alias-free output image. The main feature of this proposed error-energy reduction algorithm is that we treat the spatial samples from low-resolution images that possess unknown and irregular (uncontrolled) sub-pixel shifts as a set of constraints to populate an over-sampled (sampled above the desired output bandwidth) processing array. The estimated sub-pixel locations of these samples and their values constitute a spatial domain constraint. Furthermore, the bandwidth of the alias-free image (or the sensor imposed bandwidth) is the criterion used as a spatial frequency domain constraint on the over-sampled processing array.

The method uses multi low resolution and low dynamic imagers to incorporates the sub-pixel dynamical image formation technology with the technology of multi image reconstructing high dynamical image, and uses a special site distribution of image sensor to implement the reconstructing of ultra-high resolution image, and uses image gray level interpolation to reconstruct image gray level so as to get a high dynamic range and high resolution image.

The invention discloses a deep learning-based super-resolution image reconstruction method and system. The method comprises the steps of acquiring an image to be reconstructed and training data; inputting the training data into a multilayer convolutional neural network based on a residual structure for learning; reconstructing an optimal model acquired through input learning of the image to be reconstructed to acquire a super-resolution image. By performing deep learning through multilayer convolution based on the residual structure, the acquired optimal model can be high in super-resolution image reconstruction ability; by structuring the optimal model through a deep learning method, excessive smoothing of images acquired through an interpolation method can be avoided, and meanwhile, high-resolution images restored through the optimal model can be clear and sufficient in high-frequency details; frequency domain methods can be saved, and lack of correlation of frequency domain data canbe avoided. The deep learning-based super-resolution image reconstruction system comprises a target acquisition unit, a training unit and an image reconstruction unit and can achieve advantageous effects identical to those of the deep learning-based super-resolution image reconstruction method.

The invention discloses a super-resolution image reconstruction method based on non-local dictionary learning and biregular terms, and mainly aims to solve the problem that reconstructed images are unnatural due to the fact that prior information of ultralow-resolution images cannot be fully utilized in existing dictionary learning methods. The method includes the main steps: (1), obtaining an initial high-resolution image; (2) training an initial residual dictionary set d0 and an initial expected dictionary set D0; (3) computing an initial non-local regular weight matrix W0 and an initial local kernel regression regular weight matrix K0 on the initial high-resolution image; (4) performing regular optimization processing on an inputted initial high-resolution image to obtain an optimized image; and (5) applying the initial residual dictionary set d0 and the initial expected dictionary set D0 for reconstructing the optimized image to obtain a reconstructed image. The method is capable of reconstructing remote sensing images and effectively maintaining marginal and texture information of the images, and can be used for satellite monitoring and remote-sensing imagery.

The invention discloses a video and image lossy compression method for image coding. The method combines traditional JPEG, JPEG2000, H.264 and HEVC-code standard algorithms with super-resolution image reconstruction and designs an image compression method combining the super-resolution reconstruction on the basis. Downsampling is carried out on an input video and image, wherein a downsampling method adopts a Bicubic algorithm and a downsampling multiple is 2. The number of dot arrays of a downsampling image is only 1 / 4 of that of an original image. The encoding rate of the downsampling image is far lower than that of an original input image so that the encoding rate is reduced. At the same time, on the basis that robustness differences of a residual image and a general image are analyzed, a negative-feedback step is introduced in the design so that part of high-frequency detail information lost in a super-resolution image reconstruction step is remedied and reconstructed video or image quality is improved. Compared with the JPEG and the H.264 standard algorithms, the compression method reduces the encoding rate greatly under a situation that image quality is the same.

The invention relates to a super-resolution image reconstruction method based on analysis sparse representation, belonging to the technical field of image processing. The method comprises the following steps of: performing dictionary training according to a training sample set; and training a high-resolution dictionary and a low-resolution dictionary for an extracted feature; converting an image to be input from an RGB (Red, Green and Blue) space into a 1 alpha beta space and dividing into blocks of a same size; performing two kinds of operation on the blocks, wherein one is that each block is amplified by using the conventional amplification method and the other one is that an residual image of each block is extracted, sparse representation of the residual image in the low-resolution dictionary is calculated, and then the residual image is reconstructed in the high-resolution dictionary to obtain a reconstructed residual image; summarizing results of the two steps, converting back into the RGB space and performing back projection to obtain the reconstructed super-resolution image. According to the method, the image reconstruction noise can be obviously reduced, and detail features are kept; and meanwhile, the method has the advantages of easiness in operation and wide application.

The invention discloses a single image super-resolution reconstruction method based on image nonlocal self-similarity. The image texture is synthesized through the nonlocal self-similarity of the image and image blank information is filled; and image reconstruction is realized according to the complete theory of a deconvolution neural network. According to the super-resolution reconstruction method based on image nonlocal self-similarity convolution sparse representation, the detail information of the super-resolution image can be better enhanced, the block effect can be reduced and thus the quality of super-resolution image reconstruction can be enhanced.

The invention discloses an ultrasonic image super-resolution reconstruction method for improving contour definition based on an attention mechanism. The ultrasonic image super-resolution reconstruction method comprises the steps of S1, data acquisition; S2, network construction; S3, initializing a network; S4, network training; S5: super-resolution image reconstruction. On the basis of an existingfeature extraction reconstruction network, the method builds another level of parallel codes-codes; according to the attention mechanism network of the decoding structure, utilizing common convolution and cavity convolution, better obtaining high-frequency information in an ultrasonic image, combining the two levels of network features, and extracting the final image features by using convolutionto form a super-resolution reconstruction network. Through the two-stage parallel network, the attention mechanism network is used for positioning the specific position of the high-frequency information, the tissue interface and the tissue area in the ultrasonic image can be effectively distinguished, the edge reconstruction definition of the tissue contact surface in the ultrasonic image is improved, and the problem that the contour of the reconstructed ultrasonic image is fuzzy is solved.

The invention discloses a self-adaptive regularized super-resolution image reconstruction method which can keep marginal definition, which mainly solves the problem that the prior method has edge fog in reconstruction of a degraded image. The method comprises the following steps: an imaging model is constructed; on the basis of an unconstrained objective function constructed by a Lagrangian multiplier method, gradient is increased to approach a bound term; the objective function is expanded; L1 norm is adopted to measure a data approximation term; a self-adaptive bilateral total variation model which can carry out local adaptive control on the smoothing effect is utilized to construct a self-adaptive regular term; a gradient approximation term is added to be as constraint of gradient consistency; edge information is kept; the self-adaptive regular term and a gradient consistency bound term are introduced as constraint conditions; an expanded Lagrangian objective function is constructed and optimized; and an optimized unconstrained objective function is utilized to reconstruct an image, thereby obtaining a high-resolution image of which the edge is kept. The method can keep image edge clear, can inhibit noise and is suitable for restoration treatment on the degraded image.

The invention discloses a super resolution image reconstruction method based on gradient consistency and anisotropic regularization. The super resolution image reconstruction method based on the gradient consistency and the anisotropic regularization is used for solving super resolution image reconstruction self-adaption to maintain high-frequency image information, and recovering image detail information. The steps includes inputting a low resolution image, obtaining an interpolation image by using dual-three interpolation methods to sample the input image, adopting gradient consistency and anisotropic regularization (GCAR) conditions to restrain an objective function, performing a deconvolution operation for the interpolation image, judging a deconvoluted image whether to meet output requirements, outputting a super resolution result if the deconvoluted image meets the output requirements, otherwise, performing reconvuluting and pixel replacement for the deconvoluted image, going to a next deconvolution operation, and iterating like those until the output requirements are met. The super resolution image reconstruction method based on the gradient consistency and the anisotropic regularization has the advantages of maintaining the gradient consistency of low contrast image area low resolution images and corresponding high resolution images, and capable of recovering image detail information in a self-adaption mode and being used for the field of video applications.

The invention relates to a fine crack segmentation method based on generative adversarial networks, comprising the following steps: step one, preparing a plurality of crack images; 2, training that generator network to calculate the pixel loss; 3, training that segmentation branch of the discriminator and calculating the segmentation loss; 4, respectively reading that pixel loss and the segmentation los, on the basis of which the discriminating branch of the generator and the discriminator are jointly trained to calculate the antagonism loss; The method combines the super-resolution image reconstruction and semantic segmentation of the generated countermeasure network to design a new segmented generated countermeasure network. Compared with the traditional super-resolution image generationalgorithm, the super-resolution fine crack image quality of the invention is higher, and is more similar to the original high-resolution image.

The invention relates to visual detection and image processing. Aiming to improve the resolution of an image and the measurement accuracy of a system, the technical scheme of the invention is that: a random micro-displacement-based super-resolution image reconstruction method comprises the following steps of: calculating micro-displacement by utilizing characteristic points in images of image sequences with the random micro-displacement; performing high-accuracy extraction on coordinates of the same characteristic points; setting two original images with the pixel size of 2d, the resolution of N and the sub-pixel micro-displacement of a as A and B; and setting a to-be-reconstructed image with the resolution of 2N as H, and as the image sequences are displaced and then scenes corresponding to H0 and H1 in the image are imaged only in A and not in B, in the process of reconstructing H, making equal H0, H1 and A0, performing the reconstruction according to corresponding relationships between H2, H3 and the like, and A and B and own weights of H2, H3 and the like for H2, H3 and the like, and obtaining a formula by deduction to calculate gray values. The method is mainly applied to the image processing.

The invention provides a super-resolution image reconstruction method and device based on depth learning, including the following steps: according to the image set and the magnification of the target,a training set corresponding to the high-resolution image and the low-resolution image is established, based on the training set and the pre-constructed multi-scale network model, network training isconducted, so that the model parameter are obtained, wherein the multi-scale network model comprises a plurality of feature extraction networks and a composite network, multiple feature extraction networks are used to extract features from images, and combined networks are used to combine multiple sets of features extracted from multiple feature extraction networks; using the trained multi-scalenetwork model, the input low-resolution images are reconstructed to obtain high-resolution images. A better reconstruction effect can be obtained by extracting features from images through multiple feature extraction networks with different network depths and combining multiple features.

The invention provides a motion target super-resolution image reconstruction method based on an optical flow field. The motion target super-resolution image reconstruction method comprises the following step: first, performing motion target tracking and motion estimation based on the optical flow field; second, utilizing an inhomogeneous interpolation method to perform image fusion of low-resolution image sequences; and third, utilizing a wiener filtering method to perform image reconstruction to preliminarily-fused high-definition images to obtain clear high-definition images. In the first step, a motion target image is first captured from a first frame image, a motion target image at the same position in a next frame image is captured according to the position of a motion target image in a reference frame image, the optical flow field between the two motion target images of two frames is calculated, then motion parameters of the motion target images are obtained by utilizing the optical flow field, the positions of the motion target images in the next frame image of the reference frame image are changed according to the motion parameters, and finally adjacent frame images are performed and the motion target images of frame images are tracked or captured by means of the same method.

The invention provides a super- resolution image reconstruction system based on a self-adaptation submodel dictionary choice. The super-resolution image reconstruction system based on the self-adaptation submodel dictionary choice comprises an input module, a high and low frequency training set construction module, a candidate base vector gathering and building module, a submodel dictionary choice module, a test image preprocessing module, a super-resolution image reconstruction module and an output module, wherein the high and low frequency training set construction module comprises a band allocation submodel and a primitive block extraction submodel; the candidate base vector gathering and organizing module comprises an online dictionary learning submodel and a DCT dictionary construction submodel; the test image preprocessing module comprises a low frequency smoothing submodel and a primitive block extraction submodel. The super-resolution image reconstruction system based on the self-adaptation submodel dictionary choosing is applied to different dictionary sizes and decimation factors, the super-resolution image reconstruction system based on the self-adaptation submodel dictionary choosing can significantly improve the subjective and objective quality of a reconstitution image, the high effective dictionary design process is guaranteed, and a novel visual angle is provided for the existing image compression standard at the same time.

The invention discloses a super-resolution image reconstruction method of a generative adversarial network based on an attention mechanism, and the method comprises the steps of preprocessing an ImageNet data set, and manufacturing a training data set corresponding to a high-resolution image and a low-resolution image; constructing a generative adversarial network model for training, and introducing an attention mechanism into the model; inputting the obtained training data set into a generative adversarial network in sequence to carry out model training; and inputting the to-be-processed image into the trained generation network model to obtain a reconstructed high-resolution image. According to the invention, the attention mechanism is added into the perception network to extract the salient region of the target; a mode of combining local information and global information is utilized to enable the generated image to be closer to a real high-resolution image, and the perception lossis introduced to improve the generation effect, so that the edge and the detail information of the reconstructed image are clearer, and the reconstruction effect is better.

The invention discloses a super-resolution image reconstruction method based on sparse multi-manifold embedment. The super-resolution image reconstruction method based on sparse multi-manifold embedment comprises the steps that medium-frequency and high-frequency characteristics of a set of high-resolution training images are extracted to build a medium-frequency and high-frequency characteristic training library; clustering is carried out on the medium-frequency and high-frequency characteristic training library on the basis of the multi-manifold hypothesis, and medium-frequency and high-frequency characteristic set pairs of different classifications are obtained; medium-frequency characteristics of an input low-resolution image through the method same as the method for extracting medium-frequency characteristics of the training images, the nearest medium-frequency characteristic training center of the medium-frequency characteristics is found out, and the classification of the medium-frequency characteristic training center is appointed as a neighborhood search range of the low-resolution image; the positions of sparse neighbors, from the same manifold, of each processed medium-frequency block in the classification are determined by solving a sparse optimization problem, reconstructed high-frequency blocks are obtained through the least square solution, and after processing of all the blocks is accomplished, a high-frequency image can be formed in a composite mode; the high-frequency image is added to the amplified low-resolution image, and an initially-estimated reconstructed image is obtained; the initially-estimated reconstructed image is processed through a common post-processing method, so that the final result is obtained.

The invention relates to a remotely sensed image small object super-resolution rebuilding method, which provides a remotely sensed image small object super-resolution rebuilding model; the space resolution factor adds 1.5 ploidy of the initial picture and effect depresses the extraneous wave; the linear material and the empirical distribution estimate the atmosphere disturb image parameter H; the redundancy wavelet distribution adopts mirror-image wavelet base function and uses the cross cut correlation to achieve morphology wavelet non-linear wavelet encoding and depresses the high frequency immediately noise; it dynamically evens the effect of high frequency noise signal and high frequency detail signal. It is applied in satellite image military target identifying, small target detecting and earth source remotely sensed image measuring.

The present application provides a method and a device for super-resolution image reconstruction based on dictionary matching. The method includes: establishing a matching dictionary library; inputting an image to be reconstructed into a multi-layer linear filter network; extracting a local characteristic of the image to be reconstructed; searching the matching dictionary library for a local characteristic of a low-resolution image block having the highest similarity with the local characteristic of the image to be reconstructed; searching the matching dictionary library for a residual of a combined sample where the local characteristic of the low-resolution image block with the highest similarity is located; performing interpolation amplification on the local characteristic of the low-resolution image block having the highest similarity; and adding the residual to a result of the interpolation amplification to obtain a reconstructed high-resolution image block. The local characteristics of the image to be reconstructed extracted by the multi-layer linear filter network have higher precision. Thus, a higher matching degree can be obtained during subsequent matching with the matching dictionary library, and the reconstructed image has a better quality. Therefore, the present invention can greatly improve the quality of the high-resolution image to be reconstructed.

The invention relates to a polarization image fusion method based on super-resolution image reconstruction. The polarization image fusion method based on the super-resolution image reconstruction comprises following steps: crossing a Q image, a U image, a polarization degree image and a polarization angle image in order with a cross value being less than a spacing value of a pixel unit so as to obtain an oversampling image, and performing super-resolution reconstruction on the oversampling image so as to obtain a fusion image. By employing the polarization image fusion method, radiation intensity information and detail information such as edge contours can be effectively fused, and the resolution of the fusion image can be improved.

The invention provides a super-resolution image reconstruction method based on a progressive deep residual network, and the method mainly comprises the following steps: (1) selecting a training data set and a test data set, carrying out the rotating and scaling of a training data set image, and expanding the training data set image; (2) carrying out down-sampling processing on the obtained training data set image; (3) respectively cutting the original training data set image and the low-resolution image in the step 2 into image blocks; (4) taking the original image block and the low-resolutionimage block corresponding to the same position in the step (3) as a high-resolution / low-resolution sample pair, and generating a training data set file with a format of HDF5; (5) establishing a progressive deep residual network; (6) training a progressive deep residual network; and (7) inputting the low-resolution image into the progressive deep residual network model, and outputting to obtain areconstructed high-resolution image.

The invention discloses a super-resolution image reconstruction method based on structure self-similarity and sparse representation. The method includes the main steps of firstly, filtering a set of training sample image to extract features; then, extracting small patches to construct a dictionary including a high-resolution image block and a low-resolution image block in pair, conducting interpolation amplifying on an inputted low-resolution image, conducting filtering to extract the features, solving a reconstructed weight matrix W, conducting iteration to renew a sparse coefficient {alpha i} and a high-resolution image X to be reconstructed; finally, recovering a satisfying high-resolution image till the iteration is convergent. According to the method, the structure self-similarity of the image is used for solving the problem that an existing method is not high in quality. The operation time is short, the efficiency of image reconstruction is high, the quality of the reconstructed image is high, and various natural images which include non-texture images such as animal and plant images and human images and strong-texture images such as architecture images can be reconstructed.

The invention relates to a cluster network super-resolution image reconstruction method of an Laplace pyramid structure. By dividing the images to be reconstructed into training data sets and test data sets and enhancing and preprocessing them respectively, a Laplace pyramid cluster network is constructed and the training data sets are input for training, and the test data sets are input into thetrained Laplace pyramid cluster network to reconstruct super-resolution images step by step. A Laplace pyramid structure is adopted, By progressively reconstructing high-resolution images, Step by step optimize reconstruction results, better solve super-resolution image reconstruction, the residual learning is applied to the network, Reducing network parameters and avoiding gradient explosion caused by network complexity increase. In each construction module, there are both forward and feedback connections between any two convolution layers, and the information between layers is updated alternately, which maximizes the information flow and feedback mechanism between layers, makes the connection between layers more dense and extracts more detailed features.

Embodiments of the invention provide a recursive residual network-based super-resolution image reconstruction method. The method comprises the following step of: inputting a low-resolution image intoa trained recursive residual neural network so as to obtain a super-resolution reconstructed image, wherein the recursive residual neural network comprises a plurality of residual units, and for any residual unit, input information of the residual unit comprises output information of the last residual unit and a high-frequency feature image of a low-resolution input image. According to the method,the neural network is trained through local residual learning but not global residual learning of VDSR, so that benefit is brought to information transmission and gradient flow; non-interpolated low-resolution images are taken as inputs; and finally, super-resolution output images are directly up-sampled by using a convolution layer at the tail end of the network, and recursive structures are imported in the residual units, so that parameters are greatly decreased and the calculation complexity of the recursive residual neural network is reduced.

The invention discloses a high-precision tool setting device and tool setting method of a micro-diameter milling tool, and belongs to the technical field of mechanical automation. The device comprises a laser device, the micro-diameter milling tool, an electric spindle, a CCD chip, an X-direction high-precision sliding table, a Y-direction high-precision sliding table, a Z-direction high-precision sliding table, an image signal processing unit, a motor control unit, a laser device control unit and a computer master control unit. Z-direction high-precision tool setting of the micro-diameter milling tool at any position in the XY plane is achieved, and the laser coaxial holographic imaging technology and the super-resolution image reconstruction technology are utilized for achieving automatic measuring of a tool setting gap, so that the error caused by deformation generated by the contact between the milling tool and a workpiece during traditional try cutting is prevented, and meanwhile, the field depth error caused by the fact that the milling tool and the workpiece are not located in the same image imaging face is avoided. The device is low in cost, easy to operate, capable of being mounted in a distributed manner within the limited work space, and suitable for high-precision tool setting of a micro milling tool, a common milling tool and a numerical control milling tool.