33results about How to "Improve denoising performance" patented technology

The invention discloses an SAR image de-noising algorithm based on Primal Sketch classification and SVD domain improvement MMSE estimation. The problem that in the prior art, details are fuzzy during SAR image de-noising is solved. The method mainly comprises the steps that firstly, in the Primal Sketch algorithm, an energy image is improved by the adoption of a dual-neighbourhood contrast enhancement method, then the Primal Sketch algorithm is adopted for dividing SAR images into a non-edge class and an edge class; NLSVD decomposition is carried out on the pixel points in the two classes, a singular value matrix is estimated by the minimum mean square error criterion containing constriction factors, and inverse transformation is carried out to obtain the estimation value of the edge class and the estimation value of the non-edge class; finally, edge coefficients are calculated, and the boundary of the edge class and the boundary of the non-edge class are fused through the Butterworth fusion method to obtain a de-noising result. According to the SAR image de-noising algorithm, speckle noise in the SAR image can be effectively removed, and the edge and point target information is well kept.

The invention belongs to the field of signal processing technologies, and particularly relates to a denoising method of additive white gaussian noise signals at a low signal to noise rate. The method includes the steps that according to autocorrelation of signals, autocorrelation of the signals is solved, the autocorrelation functions of the signals obtain the maximum values at zero points, the amplitude change along with change of time difference and will not attenuate to a small value quickly; EMD is performed on the signals mixed with the gaussian white noise, due to the nature of EMD, the gaussian white noise is not the real white noise any more, however, the statistical property of the white noise approximately exists, in other words, the autocorrelation functions of all the signals mixed with the gaussian white noise obtain the maximum values at zero points, the amplitudes change along with change of the time difference and will attenuate to the small value quickly. Through the difference, IMF components of which the noise plays a main role are selected, and the influences of the noise on the signals can be effectively reduced. Under the condition of the low signal to noise rate, the denoising performance of the method is superior to that of a traditional method, and signal denoising can be completed under the condition of the low signal to noise rate.

The invention belongs to the technical field of digital image processing, provides a TV (total variation) image noise removal method based on noise priori constraint and can improve the noise detection accuracy and well protect image structural information. The technical scheme is that the TV image noise removal method based on the noise priori constraint comprises the following steps: step 1) inputting a noise-containing image I; step 2) enabling f to be an original sharp image; step 3) performing impulse noise influence on an image u in transmission and storage processes, and representing an image polluted by mixed noise with g; step 4) estimating noise positions with an ROAD (rank-ordered absolute difference) statistical method; step 5) constructing a TV-ROAD iteration noise removal model; step 6) solving an equation shown in the specification according to a noise removed image f obtained in the previous step to obtain a two-value matrix vector b. The method is mainly applied to digital image processing.

The invention discloses a transform domain non local and minimum mean square error-based SAR (Synthetic Aperture Radar) image denoising method which mainly solves the problems of edge excessive smoothness during denoising of an SAR image and difficulty in keeping point targets in the prior art. The transform domain non local and minimum mean square error-based SAR image denoising method comprises the following steps of: inputting an SAR image Y, processing one-layer non-subsample Laplace decomposition on the SAR image Y to obtain a low-frequency image YL and a high-frequency image YH; filtering the YL by using a PPB (Probalistic Patch-Based) filter to obtain a filtered image, carrying out shear wave filter decomposition on the YH to obtain a high-frequency subband image of each direction; modeling by using a Gaussian mixture model and denoising by using MMSE (Minimum Mean Square Error) estimation to obtain denoised high-frequency subband diaphragms; carrying out inverse shear wave transform on the low-frequency image YL and the high-frequency image YH to obtain a space domain image YZ; and classifying the YZ to obtain a final denoising result. The transform domain non local and minimum mean square error-based SAR image denoising method is capable of removing noise in a homogeneous region and well keeping clear edges of the images, and can be used for preprocessing the images.

The invention discloses a neighborhood adaptive Bayes shrinkage image denoising method based on a dual-tree complex wavelet domain. The method comprises the following steps: 1) performing dual-tree complex wavelet transform on a noisy image, and performing three-level decomposition to obtain multiple sub-band coefficients; 2) estimating the noise variance by use of a robust median device; 3) processing each sub-band coefficient except the low-pass sub-band coefficient in the following steps: a) calculating the variance of the noisy image in corresponding neighborhood window for each DT-CWT (dual-tree complex wavelet transform) coefficient; b) averaging the variances of the noisy image corresponding to all the coefficients to estimate the neighborhood variance of the noisy image of the sub-band; and c) assuming that a statistical model of the DT-CWT coefficients of the image obeys a GGD (general Gaussian distribution) model, estimating the optimal threshold through a minimal Bayes risk function, and softening the wavelet coefficient in the sub-band; and 4) performing dual-tree complex wavelet inverse transform reconstruction on the wavelet coefficient to obtain the denoised image. The method disclosed by the invention has perfect denoising performance and good adaptivity.

The invention discloses a fuzzy entropy-based noisy signal processing method and an iterative singular spectrum (SSA) soft threshold denoising method. The method is suitable for noisy signals. Assuming that the noisy signal of length N xin = {x1, x2, ..., xN} and assuming that the additive white noise therein is uncorrelated with the signal, a d-dimensional vector is constructed and the similarityand fuzzy probability are defined by utilizing an original signal xin; a (d + 1)-dimensional vector is constructed and the corresponding similarity and fuzzy probability are defined by the same method; and the fuzzy entropy is defined in the drawing of the description. For components obtained by utilizing a known signal decomposition method, the singular spectrum distribution of all the components is defined as a fuzzy entropy spectrum. The fuzzy entropy for quantifying the complexity of the system in a chaos theory is utilized to characterize a noise plane and provide a more effective path for the processing of the noisy signal; the fuzzy entropy spectrum-based iterative singular spectrum (SSA-IST) soft threshold denoising method has the denoising performance better than that of the traditional truncated singular spectrum method, and wavelet transform and empirical mode decomposition denoising method.

The invention discloses a rolling bearing fault diagnosis method based on vibration signal denoising and envelope analysis and belongs to the field of mechanical fault diagnosis and signal processing.The method comprises the following steps: firstly, performing primary collaborative filtering and denoising processing on four types of vibration signals of an outer ring fault, an inner ring fault,a rolling body fault and a normal condition of a rolling bearing, then performing primary collaborative filtering and denoising processing once again on the signals subjected to collaborative filtering and denoising, further removing noise interfering with fault diagnosis in the signals, afterwards, performing empirical mode decomposition (EMD) on the signals subjected to secondary collaborative filtering and denoising to obtain a plurality of implication mode functions (IMF), and selecting an IMF1 to carry out envelope analysis to judge the fault type of the rolling bearing. The result showsthat the vibration signals can be effectively denoised after being subjected to secondary collaborative filtering, then, envelope analysis based on empirical mode decomposition is carried out on the denoised vibration signals, and fault diagnosis can be accurately and effectively carried out.

The invention discloses a rank reduction and denoising method for singular value attenuation. The method is applied to the seismic data processing field. In the prior art, noisy seismic data cannot bewell decomposed into a noise subspace and a signal subspace, the effect is not ideal under the condition of low signal-to-noise ratio, for the above problem, the invention provides the method which comprises the following steps: firstly, calculating optimized weight coefficients of left and right singular vectors of a noisy observation matrix; using the characteristic that an optimized weight coefficient only depends on limit singular value distribution of a pure noise matrix to decompose the noisy signal well into the noise subspace and the signal subspace, then adding a regularization operator to constrain the singular value on the above basis so as to obtain more stable low-rank estimation, and finally, obtaining stable estimation of seismic data in a frequency domain by utilizing reverse diagonal average processing. The method has outstanding denoising capability in the application of suppressing seismic random noise.

The invention discloses an image denoising method combining a wavelet packet and a partial differential equation. The method comprises the following steps of carrying out grayscale conversion on a collected original image and carrying out noise adding processing; constructing a new denoising model combining a PM model and a MCD model, through selecting a weight function, distinguishing and isolating noises, and then establishing a second-order differential operator; establishing the image denoising method based on the wavelet packet and the partial differential equation, using the wavelet packet to carry out coefficient decomposition on a noise image, and centrally processing noise information; according to a wavelet packet decomposition coefficient, using wavelet packet inverse transformation to reconstruct the image; and finally, carrying out smoothing processing on the processed image, carrying out simulation through a semi-implicit additive operator split numerical algorithm, comparing a mean square error, a peak signal to noise ratio and definition, and analyzing the validity and the feasibility of the method. By using the method, the noises are effectively removed, and simultaneously, the edge texture and other details of the image can be protected, and denoising performance is excellent.

The invention belongs to the technical field of radars, and particularly relates to a secondary radar signal denoising processing method based on a deep four-channel network. According to the specificscheme, firstly, a secondary radar response signal data set is prepared, and data set division and data set preprocessing are carried out; and then a novel deep four-channel neural network is constructed based on deep learning, wherein the novel deep four-channel neural network is formed by connecting four neural branches in parallel; the four paths can fully fuse learned different deep signal features, so that feature loss is reduced; and finally, the secondary radar test data are input into the deep four-channel neural network to suppress noise, and timing response signals are predicted. The method has very high denoising performance, and meets the requirement of noise suppression when the secondary radar receives the response signal.

The invention relates to the field of network security, and discloses a network intrusion detection method based on element sample sparse representation. The problems that a traditional common intrusion detection method is difficult to deal with unknown network attacks and the false alarm rate of anomaly detection is high are solved. The method comprises the steps that a, a network data collection tool is used to collect network instances to build a training sample set; b, an element sample is extracted from the built network training sample set in a centralized manner, and an element sample set is formed to replace the training sample set; and c, a sparse representation classification method is used to detect network data to be identified, so as to identify the category of the network data to be detected. The network intrusion detection method provided by the invention is applicable to network intrusion detection.

The invention belongs to the technical field of radars, and particularly relates to a secondary radar signal denoising processing method based on a deep residual separation convolutional network. Themethod comprises the steps of firstly, preprocessing a secondary radar response signal, and carrying out data set division, normalization and vectorization on data; and then, based on a deep learningmethod, constructing a deep residual separation convolutional neural network which comprises three parts of shallow feature extraction, down-sampling deep feature extraction and up-sampling feature fusion; adopting separable convolutional neural network connected by the deep residual error to effectively extract the deep features of the secondary radar signal; and finally, inputting the data intothe network to predict a secondary radar response signal. The method has very high denoising performance in a normal working environment of the secondary radar, can accurately predict the secondary radar response signal, and ensures correct decoding of the secondary radar signal.

The invention discloses a low-light image enhancement method based on noise attention map guidance, and the method comprises the steps: 1, constructing a noise estimation module which inputs an original low-light-level image, wherein the size of the original low-light-level image is H * W * 3, and the output of the noise estimation module is a characteristic pattern with the size of H * W * 1; and 2, constructing an enhanced network module, wherein the input data of the enhanced network module are the output features in the step 1 and the original low-light image, the size of the enhanced network module is H * W * 4, the output of the enhancement network module is an enhanced image, the size of the enhanced image is H * W * 3, and the enhanced network module comprises an encoder, a ResidualBlock and a decoder. The method provided by the invention can effectively enhance a single low-light image, has good denoising performance under the guidance of a noise attention graph, does not cause color distortion, can retain texture details of an original graph, and has good generalization ability on different data sets.

The invention relates to an L0 norm-based three-dimensional grid denoising method. The method comprises the following steps of obtaining edges and vertexes of a three-dimensional grid; setting an optimization goal according to the obtained edges and vertexes of the three-dimensional grid; and introducing a laplace operator for optimization of the three-dimensional grid according to the optimization goal to obtain the denoised three-dimensional grid. Compared with the prior art, the L0 norm-based three-dimensional grid denoising method has the advantages of being fast in calculation, high in efficiency, convenient to calculate, smooth to implement and the like.

The invention relates to the field of rolling surface defect detection, in particular to a rolling surface defect detection system and method. The rolling surface defect detection system comprises a control device, a light source generation device, a galvanometer and an identification device, wherein the galvanometer is arranged on a light path between the light source generation device and a rolling surface to be detected; the light source generating device, the galvanometer and the recognition device are all in communication connection with the control device; light beams emitted by the light source generating device are projected to the rolling surface to be measured through scanning of the galvanometer to form stripes; the control device adjusts driving signals of the galvanometer to adjust the positions of the stripes projected to the rolling surface to be measured; and the identification device shoots the rolling surface to be detected to obtain multiple target images and detectsthe defects of the rolling surface to be detected according to the multiple target images. The detection accuracy rate of the defects of the rolling surface to be detected can be effectively improvedby detecting and processing the target images shot multiple times, meanwhile, the system is high in detection precision, the detection process is simple and easy to operate, and high market value isrealized.

The invention discloses a data acquisition and processing system of an elliptic curved crystal spectrometer. The system comprises a CCD photoelectric conversion module, an AD application module at ananalog front end, an FPGA driving and controlling module, a USB interface transmission module and a system power supply module. The system obtains a spectral image of a laser plasma X-ray, realizes data communication with a PC through the USB interface transmission module, and displays the transmitted data on the PC. The system can acquire image data with different frame frequencies and differentresolutions in real time and at high speed, and transmit and display the acquired data in real time, and the system obtains a higher signal-to-noise ratio and a good visual effect for the spectral image by using wavelet transformation and a median filtering algorithm. The system obtains a spectral diagram of the X-ray by using the denoised spectral image, and converts the relation between pixels and light intensity into the relation between wavelength and intensity through wavelength calibration, so that the spatial resolution of the elliptic curved crystal spectrometer is obtained, and the temperature and the rotating speed of the plasma are obtained.

The invention discloses a real image denoising method based on a multi-scale selection feedback network. The method comprises the following steps: constructing a multi-scale selection module MSB for extracting multiple receptive field scale features; constructing a multi-scale selection feedback network MSFB, wherein the MSFB comprises a shallow feature extraction unit, a plurality of MSBs connected in series, an image reconstruction unit and a degradation model; for image denoising, two dual tasks are constructed: predicting a noise-free image from an original noise image, and degrading the predicted noise-free image to a noise image; repeatedly executing two dual tasks in a plurality of time steps by using the MSFB, and performing multi-stage iteration; in iteration, selectively feeding back high-level semantic information output by the deep MSB of the previous time step to the input end of the shallow MSB of the next time step, and the MSFB is trained through iteration; the training process taking minimization of dual loss as an optimization target and taking a peak signal-to-noise ratio as an evaluation index of network performance; and inputting a noise image into the trained MSFB for de-noising, and outputting the de-noised image.

The invention discloses a transform domain non local and minimum mean square error-based SAR (Synthetic Aperture Radar) image denoising method which mainly solves the problems of edge excessive smoothness during denoising of an SAR image and difficulty in keeping point targets in the prior art. The transform domain non local and minimum mean square error-based SAR image denoising method comprises the following steps of: inputting an SAR image Y, processing one-layer non-subsample Laplace decomposition on the SAR image Y to obtain a low-frequency image YL and a high-frequency image YH; filtering the YL by using a PPB (Probalistic Patch-Based) filter to obtain a filtered image, carrying out shear wave filter decomposition on the YH to obtain a high-frequency subband image of each direction; modeling by using a Gaussian mixture model and denoising by using MMSE (Minimum Mean Square Error) estimation to obtain denoised high-frequency subband diaphragms; carrying out inverse shear wave transform on the low-frequency image YL and the high-frequency image YH to obtain a space domain image YZ; and classifying the YZ to obtain a final denoising result. The transform domain non local and minimum mean square error-based SAR image denoising method is capable of removing noise in a homogeneous region and well keeping clear edges of the images, and can be used for preprocessing the images.

The invention relates to an image noise reduction method and device, equipment and a storage medium, and the method is used for the noise reduction of a thermal image of electrical equipment, and comprises the following steps: obtaining a to-be-processed thermal image of the electrical equipment; performing waveband decomposition on the thermal image of the electrical equipment to obtain a plurality of waveband subsets; introducing a Gaussian noise item and a sparse noise item, and constructing a hierarchical dictionary learning model for each waveband subset; and solving the constructed hierarchical dictionary learning model by adopting a Metropolis-Hastings algorithm, so as to obtain a noise reduction image. Compared with the prior art, the method has the advantages of good denoising performance and the like.

The invention belongs to the technical field of radars, and particularly relates to a secondary radar signal processing method based on a two-channel residual error deep neural network. The method comprises the following steps: firstly, acquiring secondary radar response signal sample data and preprocessing a data set; then constructing the novel two-channel residual error deep neural network based on a deep learning method, the two-channel residual error deep neural network is composed of two feature extraction channels, each channel is subjected to residual error addition for multiple times,and residual error connection is carried out between the two channels; inputting the training set and the verification set, training the residual two-channel deep network, and stopping training whenthe parameters are optimal; and finally, inputting the test data into the network, and predicting a secondary radar response signal. The network model can reduce information loss and fully extract deep features of secondary radar signals. The method is excellent in denoising performance, can accurately predict the secondary radar time sequence signal, and meets the noise inhibition demands.

The invention discloses a rolling bearing fault diagnosis method based on CEEMDAN and GWO-NLM, and the method comprises the steps: carrying out the decomposition of a fault signal collected by a sensor through employing CEEMDAN, screening the decomposed IMF components through employing a correlation coefficient-energy ratio-kurtosis criterion, and reconstructing the screened signal into a new signal; performing parameter optimization selection by using a non-local mean filter optimized by a grey wolf algorithm to realize an optimal denoising effect, performing SG smooth filtering on the denoised signal to perform secondary denoising, and performing feature extraction on the finally obtained signal to obtain fault features of the bearing; and finally judging the fault type of the rolling bearing according to the fault characteristics of the bearing. According to the method, noise interference in the vibration signal can be effectively suppressed, and the fault diagnosis accuracy of the bearing is improved.

The invention belongs to the technical field of medical imaging, and discloses a mathematical algorithm microscopic bladder endoscope imaging system and an image processing method. The system comprises an endoscope handle, an outer sheath tube, an image processing module and a display module; the endoscope handle is connected with the outer sheath tube, the outer sheath tube is provided with a bending part, a direction control mechanism is connected with the bending part, and the direction control mechanism is connected with the image processing module; the end, away from the outer sheath tube, of the bent portion is provided with a tip portion, an image sensor is arranged in the tip portion, the light source module provides a light source for the image sensor, the image sensor is connected with the image receiving module, the image receiving module is connected with the image processing module, and the image processing module is connected with the display module. The method comprises the steps that the light source module provides a light source for the image sensor, controls the angle of the bending part, collects an image in the body of a patient, carries out image processing, and displays image information in the body of the patient on the display module. According to the invention, the quality of medical images is improved, and research and judgment of tumor risks are realized. The device is suitable for medical staff to observe conditions in the body of a patient.

The invention discloses a baseband non-stationary digital signal denoising method, a baseband non-stationary digital signal denoising system and a storage medium thereof, and aims to solve the technical problem of relatively high calculation complexity of a denoising method in the prior art. The method comprises the following steps: designing a frequency domain filter for the characteristics of abaseband non-stationary digital signal; performing fast Fourier transform on the to-be-processed baseband non-stationary digital signal to obtain a frequency spectrum of the to-be-processed signal; multiplying the frequency spectrum of the signal to be processed by a frequency domain filter, and carrying out frequency domain filtering; and performing inverse fast Fourier transform on the frequencydomain filtering result to obtain a denoised output signal. The method has a good denoising effect, and compared with the prior art, the method is low in calculation complexity and higher in calculation speed.

The invention discloses a microphone array speech enhancement method based on rear FastICA signal separation. The method comprises the steps that microphone array element receiving signals are acquired and preprocessed; a fixed beam former of a upper branch performs primary noise reduction on the microphone array element receiving signal; an adaptive blocking matrix of a lower branch filters a desired signal, and an adaptive canceller outputs multiple paths of noise signals as one path of noise signal; and a FastICA algorithm is performed on the output signals of the upper and lower branch adaptive cancellers to separate voice and noise, and an enhanced voice signal in the separated signals is obtained. According to the method, leakage of target signals can be reduced, the signal-to-noise ratio of voice can be improved and enhanced, and especially when the signal-to-noise ratio of background noise Gaussian white noise is too low, the method can show excellent denoising performance and is higher in robustness.

The invention discloses an image denoising method and device based on Jacobian dynamic approximation, and relates to the technical field of computer vision. Comprising the steps of obtaining a to-be-denoised image; inputting an image to be denoised into a shallow feature extraction unit of the image denoising network to obtain a shallow feature image; inputting the shallow feature image into a denoising unit of an image denoising network to obtain a denoised image; and inputting the denoised image into a double-attention unit of an image denoising network to obtain a final denoised image. According to the method, an end-to-end image denoising network based on Jacobian dynamic approximation is constructed, and an ordinary differential equation solving module based on a Jacobian matrix is designed to realize forward derivation of a neural network. The method can be successfully applied to image denoising of synthetic noise and real noise, and compared with a classic image denoising method, the method has the advantage that the denoising performance can be obviously improved in the subjective and objective quality aspects of a reconstructed image.