32 results about "Singular value thresholding" patented technology

The invention provides an image repair method and an image repair device. The image repair method includes determining at least one missing block containing unknown pixel points in an image to be repaired; for any missing block among the at least one missing block, clustering to obtain a similar block of the any missing block, so as to form a category obtained by clustering of the any missing block; forming a matrix having low-rank property according to the category obtained by clustering of the any missing block; adopting a nuclear norm to approach the rank of the matrix, and adopting a singular value threshold method to solve the matrix; and repairing any missing block in the image to be repaired according to the value after solving the matrix. The technical scheme of the invention improves accuracy of repairing the missing blocks in the image, and effectively achieves the purpose of repairing the image with the missing blocks.

The invention discloses a super-resolution image reconstruction method and system based on multi-feature learning, and the method makes full use of the rich information contained in a single input image for reconstruction, and does not depend on an external database. According to the method, a mapping relation between image features is established based on cross-scale similarity of images, and a high-resolution image containing high-frequency information is reconstructed for an input image directly by using the mapping relation, so that the defect of high-frequency information loss caused by image reconstruction by using an interpolation amplification method is well overcome. According to the method, effective high-frequency information is acquired by using singular value thresholding, andthe high-frequency information is amplified by using a gradient feature mapping relation and then is overlapped on a high-resolution image in a blocking manner, so that a final image reconstruction result is obtained. According to the method for reconstructing the image by utilizing the image feature combination, noise points of the reconstructed image are effectively inhibited, image edge and texture information is well kept, and detail enhancement of the image is realized.

The invention discloses a low-rank structure-based sparse compressive sensing MRI (Magnetic Resonance Imaging) image reconstruction method, which mainly solves the problem of difficulty in accurate recovery of an MRI image existing in the conventional technology. The method comprises the following implementation steps: initially recovering the MRI image by using the conventional compressive sensing and looking for a similar block matrix from the image to form an index set; performing singular value decomposition on the similar block matrix and calculating a threshold, and performing threshold calculation on a singular value by using the threshold to obtain an after-threshold singular value; and optimizing the MRI image by using the after-threshold singular value, i.e., circularly performing processes of updating of the similar block matrix and the index thereof, similar block matrix singular value decomposition and threshold and singular value threshold calculation on the MRI image to obtain a final recovery image. The recovered MRI image is clearer and has sharper edges; and the method can be used for processing a medical image.

A dynamic video segmentation method based on weighted non-convex regularization and iteratively re-constrained low-rank representation, including the following steps: (1) introduce a weighting factor W for the error matrix, and determine the constraint form of the weight matrix; (2) combine the W matrix to calculate the space Laplacian structure matrix L; (3) introduce weighted non-convex Rational function for the singular value of representation matrix Z; (4) optimize the existing GLRR framework through steps 1, 2 and 3, and propose the IRWNR model; (5) adopt the IRM framework (algorithm 1) Iteratively optimize the unknown variables W, L and Z in the target model; (6) Use the proximal gradient (EIPG) algorithm (Algorithm 2) to solve the subproblem of Z; (7) Use the block singular value threshold approximation method (Algorithm 3) Realize the SVT operation in Algorithm 2; (8) Iteratively optimize to obtain W, L and Z, which are used to realize dynamic video segmentation. It has the advantages of high operating efficiency, strong data adaptability, high accuracy and strong robustness.

The invention discloses a light-excited infrared thermal imaging defect detection method based on structured sparse decomposition. By performing wavelet decomposition on each frame image of the heat map sequence, only the low-frequency part is reserved to form a new heat map sequence and arrange them sequentially. Restructured into a new matrix, the new matrix is ​​decomposed into a low-rank matrix, the sum of a sparse matrix and a noise matrix, where the low-rank matrix represents the background of the thermal image, and the sparse matrix represents the defects in the thermal image. The sparse matrix is ​​further decomposed into the product of a dictionary matrix and a weight matrix, where the dictionary matrix is ​​used to characterize the different thermal modes of different defects on the same specimen, and the weight matrix has sparse constraints and non-negative constraints. The low-rank matrix is ​​solved by the singular value threshold decomposition method, the dictionary matrix is ​​solved by the vertex component analysis method, the weight matrix is ​​solved by the multiplier alternating direction method, and finally the sparse matrix is ​​reconstructed into a defect image matrix, so as to realize the defect detection of infrared thermal imaging .

The invention relates to a channel estimation method of a frequency domain filtering mobile WiMAX (world interoperability for microwave access) downlink system based on singular value decomposition, which comprises the following steps: using the least-square estimation method to get channel frequency response estimation of an actually distributed pilot symbol subcarrier in each OFDM (orthogonal frequency division multiplexing) symbol; getting the channel frequency response estimation of the subcarriers of the adjacent OFDM symbols by channel frequency response linear interpolation in the corresponding subcarriers in the adjacent OFDM symbols; merging the actual pilot subcarriers with the channel frequency response estimation got by linear interpolation, arranging index values of all the subcarriers in an ascending order, and then sending into a frequency domain filter for denoising after judging a singular value threshold; and finally re-utilizing the linear interpolation to get the final channel frequency response estimation of all the data subcarriers. By using the frequency domain filter, the impacts of noise in a frequency domain least-square channel estimator can be reduced; by performing the judgment of the threshold based on the singular value decomposition on the frequency domain filter, the stability of frequency domain filtering is increased, and the problem of edge effect caused by non-uniform distribution of the pilot symbols is overcome.

Provided is a temperature field data reconstruction method. The temperature field data reconstruction method includes the steps of 1, obtaining a temperature field data matrix; 2, conducting random sampling on the temperature field data matrix; 3, on the basis of sample data, filling the temperature field data matrix to reconstruct temperature field data. Before random sampling is conducted on thetemperature field data matrix, rank reduction processing can be conducted on an original matrix by adopting such a singular value decomposition algorithm. A singular value threshold algorithm or a singular value projection algorithm can be adopted to fill the temperature field data matrix on the basis of the simple data to reconstruct the temperature field data. By the adoption of the temperaturefield data reconstruction method and a computer readable storage medium, the problem is solved that in traditional methods, only mathematic relations among stations or correlations among locals are taken into consideration; therefore, global temperature field data can be recovered by using little temperature field station data; at the same time, reconstruction errors are low, the signal-to-noiseratio is extremely high, and not only is a traditional interpolation method overcome, but also data storage space can be saved.

The invention discloses a thick cloud remote sensing removal method and device and belongs to the technical field of image processing. The thick cloud remote sensing removal method comprises the following steps of s101, acquiring a reliable element set from a matrix of images to be recovered through deterministic sampling; s102, setting step length and soft threshold parameters according to a singular value and threshold value shrinkage iterative algorithm; s103, performing shrinkage iteration through the singular value and threshold value shrinkage iterative algorithm to reconstruct an original image matrix; s104, judging whether a relative reconstruction error is smaller than the preset error; s105, adding one time on iterations; s106, judging whether the number of iterations is the largest; s107, outputting the reconstructed image matrix. By means of the thick cloud remote sensing removal method, defect information caused by pollution, damage, blocking or the like can be repaired well, information can be recovered, detail texture information is remained, and consistency of image structures is kept.

The invention discloses a moving target detection method based on non-convex low-rank sparse decomposition, the steps are: 1. 0 The non-convex surrogate function in the general form of norm is applied to the approximation of low-rank matrices; and the non-convex surrogate function with good performance is selected; 2. The non-convex low-rank sparse decomposition model is established; 3. The singular value threshold SVT uses the generalized singular value The threshold GSVT is used instead; 4. Use the alternating direction multiplier method to solve the non-convex low-rank sparse decomposition model, and obtain the foreground target according to the obtained solution. The invention solves the problem that the approximate expression of the rank function in the existing method is inaccurate, resulting in that the recovered background is not clean enough and the accuracy of the extracted foreground target decreases.

The invention relates to a method for estimating the arrival rate of an intersection based on sampled trajectory data, comprising the following steps: 1) obtaining sampled vehicle trajectory data, and preprocessing the arrival rate between queuing trajectories into an initial arrival rate matrix; 2) according to the initial Arrival rate matrix, filling the empty values ​​in the initial arrival rate matrix through the singular value threshold algorithm to obtain the filled arrival rate matrix; 3) Correcting and repairing the filled arrival rate matrix to obtain the final arrival rate matrix. The arrival rate vector of each period is accumulated and multiplied by the length of the unit time interval to obtain the flow rate of each period. Compared with the prior art, the present invention has the advantages of strong generality, good expansibility, wide applicability and the like.

The invention relates to a periodic long-short code direct sequence spread spectrum code division multiple access signal multi-pseudo-code estimation method. The existing direct sequence spread spectrum code division multiple access signal pseudo code blind estimation technology cannot be applied to periodic long-short code direct sequence spread spectrum code division multiple access signals adopting short code spread spectrum and long code scrambling. The method comprises the steps of: constructing periodic long-short code direct sequence spread spectrum code division multiple access signals with complicated structures into a deficit matrix model of short code direct sequence spread spectrum code division multiple access signals, and modeling compound code matrix estimation as a blind source signal separation problem; applying a matrix filling theory to the compound code matrix estimation, and estimating user compound code sequences based on a singular value threshold algorithm and a fast independent component analysis algorithm; and finally proposing a delay triple correlation algorithm by utilizing a displacement stacking feature of an m sequence, and estimating long-short pseudo code sequences contained in the user compound code sequences. The periodic long-short code direct sequence spread spectrum code division multiple access signal multi-pseudo-code estimation method fully utilizes the matrix filling mathematic model and the triple correlation peak feature of the m sequence, and achieves blind estimation of the user compound code sequences, a long scrambling code sequence and a short spread spectrum code sequence of the signals.